Ching-Ming’s Work Log 10/11/2004-10/17/2004 10/18/2004-10/24/2004 10/25/2004-10/31/2004 11/01/2004-11/07/2004 11/08/2004-11/14/2004 11/15/2004-11/21/2004 11/22/2004-11/28/2004 11/29/2004-12/05/2004 12/06/2004-12/12/2004 12/13/2004-12/19/2004 12/20/2004-12/26/2004 12/27/2004-01/02/2005 01/03/2005-01/09/2005 01/10/2005-01/16/2005 01/17/2005-01/23/2005 01/24/2005-01/30/2005 01/31/2005-02/13/2005 02/14/2005-02/20/2005 02/21/2005-02/27/2005 1 Ching-Ming’s Work Log 01/05/2004-01/11/2004 01/12/2004-01/18/2004 02/02/2004-02/08/2004 02/09/2004-02/15/2004 02/16/2004-02/22/2004 02/23/2004-02/29/2004 03/01/2004/03/07/2004 03/08/2004-03/14/2004 03/15/2004-03/21/2004 03/22/2004-03/28/2004 03/29/2004-04/04/2004 04/05/2004-04/11/2004 04/12/2004-04/18/2004 04/19/2004-04/25/2004 04/26/2004-05/02/2004 05/03/2004-05/09/2004 05/10/2004-05/16/2004 05/17/2004-05/23/2004 05/24/2004-05/30/2004 05/31/2004-06/06/2004 06/07/2004-06/13/2004 06/14/2004-06/20/2004 06/21/2004-06/27/2004 06/28/2004-07/04/2004 07/05/2004-07/11/2004 07/12/2004-07/18/2004 07/19/2004-07/25/2004 07/26/2004-08/01/2004 08/02/2004-08/08/2004 08/09/2004-08/15/2004 08/16/2004-08/22/2004 08/23/2004-08/29/2004 08/30/2004-09/05/2004 09/06/2004-09/12/2004 09/13/2004-09/19/2004 09/20/2004-09/26/2004 09/27/2004-10/03/2004 10/04/2004-10/10/2004 2 Semester Plan (2004-Q3) Task NBL Research Month Subtask Accton Cooperation Recruit new blood Job transfer VIA Networks test Hawking Tech. test Submit Shuo-Yen’s Paper Web QoS algorithm Web QoS simulator 3 October 10 17 24 31 7 November 14 21 28 5 30 January 16 23 9 2 December 12 19 26 Finish Date 2/28/2005 11/31/2004 1/31/2005 11/31/2004 12/15/2004 10/15/2004 12/31/2004 Remark IEEE IC 3 Half-Year Plan (2004-Q1Q2) Task NBL Month Subtask B&R testbed construction Intern training program B&R testing service Part-time job at Accton Qualifying exam Operating System Artificial Intelligent WebQoS Algorithm Simulator developing Preliminary result Complete result Paper writing Paper Task NBL WebQoS E-commerce evaluation CDN Month Subtask B&R testing service Part-time job at Accton Preliminary result Complete result Paper writing 4 January 11 18 25 February 15 22 1 8 29 7 March 14 21 28 4 April 11 18 25 2 9 May 16 23 2/15 4/15 5/15 6/30 8/31 1/25 1/25 30 6 June 13 20 27 Remark 2/6 2/6 Finish Date 1/31 2/28 Finish Date J: ECR J: IEEE IC Remark 5/15 6/30 8/30 4 Works Done Last Week (12/29/2003-01/04/2004) NBL: (*)1. Bridging and routing panel (*)2. Phone or visit vendors who cannot attend the panel discussion STP is done. RTSP is still on going. 5. Track the purchasement of Bridging and Routing equipment Done. (*)4. Finish the slides for spanning tree protocol and rapid spanning tree protocol Ambit is done. 建漢 and 上元 are still on going. (*)3. Interview the applicants for the intern job Done. I have told the definite requirement to 彰燦. Coursework: 1. Prepare for the AI final On-going. 2. Prepare for the computer networks final On-going. 5 Works Planned for This Week (01/05/2004-01/11/2004) NBL: (*)1. Meet with Accton to discuss the cooperation project (*)2. Phone or visit vendors who cannot attend the panel discussion (*)3. Finish the training materials for rapid spanning tree protocol (*)4. Meet with the interns to discuss the part-time job 5. Buy remote controllers and terminal servers Coursework: (*)1. Prepare for the AI final (*)2. Prepare for the computer networks final 6 Description (01/05/2004-01/11/2004) I have met with Mgr. Weng and Mgr. Chou to discuss the part-time job at Accton. The job at Accton is to design test cases, execute designed cases, and write down all the detailed testing procedures. The purpose of the regular meetings at NBL for this project is to work out possible combinations of features for testing, purpose for each test case, and possible testing procedures. That is, before fulfilling testing procedures at Accton, we should think how to create test cases in our regular meetings. So the regular meetings help creating practical testing procedures at Accton, not part-time job at Accton helps designing test cases in our regular meeting. They asked me to tell them the following things by the end of this week: 1. Staff information (Who, When to begin, working time-slots, and pay) 2. Feature matrix (A matrix showing the combinations of the target features) 3. Project finish date (When this project will be finished) 7 Works Done Last Week (01/05/2004-01/11/2004) NBL: (*)1. Meet with Accton to discuss the cooperation project (*)2. Phone or visit vendors who cannot attend the panel discussion Done. 5. Buy remote controllers and terminal servers Done. (*)4. Meet with the interns to discuss the part-time job Done. But CyberTan and ADMTek did not respond. (*)3. Finish the training materials for rapid spanning tree protocol Done. I found some retailers in Taiwan. I will buy them this week. Coursework: (*)1. Prepare for the AI final Done. (*)2. Prepare for the computer networks final Done. 8 Works Planned for This Week (01/12/2004-01/18/2004) NBL: Begin to part-time at Accton Send panel summary to the participants Begin to train the intern students Buy remote controllers and terminal servers Pass the Accton part-time message to the applicants Research: Report WebQoS in Thursday’s meeting 9 Description (01/12/2004-01/18/2004) I will begin the part-time job at Accton this week. I will ask Accton to let me go through their test plans in the initial stage so as to know their testing procedures, testing tools and possible problems. This can help us to design test cases in the future. The part-time students will begin their work from mid-February. Before that, I will give them some training in the background knowledge and hands-on skills of bridging and routing. I have collected many training materials including Cisco CCNA/CCNP training materials, e-books, IEEE standards, IETF RFCs and testing methodologies. I think we can learn a lot if the training is well planned. 10 Works Done Last Week (01/12/2004-01/18/2004) NBL: Begin to part-time at Accton Send panel summary to the participants Su-Yi is asking the local agents The racks and workbenches will be shipped to NBL this week Pass the Accton part-time message to the applicants The training materials have been uploaded to the website Buy remote controllers and terminal servers done Begin to train the intern students I have setup a working environment done Research: Report WebQoS in Thursday’s meeting done 11 Works Planned for This Week (02/02/2004-02/08/2004) NBL: Teach the intern students to use testing equipment Write the bridging and routing testing items Accton: perform the spanning tree protocols conformance testing Track the purchase of the racks and other equipment Setup a bridging and routing testing environment Qualifying exams: Operating system Artificial Intelligence 12 Description (02/02/2004-02/08/2004) I report my observation on Accton’s current testing capabilities here. For bridging testing, they leverage IOL’s test suites and have designed their own test plans for the lacked parts (e.g. IEEE 802.1X). They have a team dedicated to developing automatic test scripts according to their test plans. However, the quality of the test scripts depend on the test plans. They are reviewing the test plans again to improve the quality and try to find out the possible combinations of functionality testing. For the routing testing, they do not make too many efforts on it now. So they test routing functions manually if not using ANVL. But their ANVL is outdated and has many bugs. I think there is still a long way for them to go. 13 Works Done Last Week (02/02/2004-02/08/2004) NBL: Teach the intern students to use testing equipment Write the bridging and routing testing items Ben asked me to retrieve the intelligent hub though express delivery. Setup a bridging and routing testing environment I have finished 50% of the STP conformance testing and got many failures in the timer testing. Track the purchase of the racks and other equipment I have finished the interoperability test items. Accton: perform the spanning tree protocols conformance testing I have taught Ch1-3 of the training materials and switch configuration. The racks have been sent to R701 but not assembled. Qualifying exams: Operating system Done Artificial Intelligence Done 14 Works Planned for This Week (02/09/2004-02/15/2004) NBL: Teach the intern students routing and router configuration Write the bridging and routing testing items Accton: perform the spanning tree protocols conformance testing (STP, RSTP) Track the purchase of the racks and other equipment Setup a bridging and routing testing environment Research: Revise the e-commerce and CDN papers 15 Description (02/09/2004-02/15/2004) Last week I started to test the spanning tree protocol (STP) conformance of the Accton L3 switch. Before the testing, I thought the pass ratio should be 100% because STP is a classical protocol. However, the test results shown many failures in the timer-related test cases. I found the timer of the DUT is very inaccurate even the configured time is very short. The inaccurate timer would affect many operations of STP, such as MAC entry expiration , port state transition and topology change. I will report the result to Accton and ask why this happened in their latest product. 16 Works Done Last Week (02/09/2004-02/15/2004) NBL: Teach the intern students routing and router configuration Write the bridging and routing testing items I got a quotation from a local vendor but the list price is very high. Therefore, I prefer to but the devices from the Internet. Setup a bridging and routing testing environment The pass ratio of STP testing is 41.5%. The RSTP testing is under testing. Track the purchase of the racks and other equipment Version 0.1 was done but withgout point count and CDRouter testing. Accton: perform the spanning tree protocols conformance testing (STP, RSTP) I have taught basic testing methodologies and IxANVL. Done. Research: Revise the e-commerce and CDN papers The e-commerce paper was done. I will submit it soon. 17 Works Planned for This Week (02/16/2004-02/22/2004) NBL: Intern student: perform the STP conformance testing Add CDRouter testing to the bridging and routing test items Accton: perform the spanning tree protocols conformance testing (RSTP, MSTP) Order remote power controllers and a terminal server from the Internet Research: Submit the e-commerce paper and revise the CDN paper 18 Description (02/16/2004-02/22/2004) NBL hired some intern students to help the bridging and routing testing and to take off the load from boring repeated jobs. In order to improve the performance of the intern students, effective team management is needed. Some management strategies such as job assignment, progress tracking, bi-weekly report, biweekly meeting, and monthly meeting are brought into action. I think these can be great helps. However, we need some time to see the results. Besides, exchanging management experiences with the other project managers is another great way. 19 Works Done Last Week (02/16/2004-02/22/2004) NBL: Intern student: perform the STP conformance testing Add CDRouter testing to the bridging and routing test items 2/3 of RSTP testing was done. Order remote power controllers and a terminal server from the Internet Done. Accton: perform the spanning tree protocols conformance testing (RSTP, MSTP) Partially done. I have asked for a quotation from the agent of DigiTech, but the list price is high. Research: Submit the e-commerce paper and revise the CDN paper The e-commerce paper was submitted to Electronic Commerce Research. The CDN paper is till under revising. 20 Works Planned for This Week (02/23/2004-02/29/2004) NBL: Intern student: finish the STP conformance testing Accton: perform the spanning tree protocols conformance testing (RSTP, MSTP) Ask for a new quotation of the PDU and terminal server Discuss the testing program for the ADSL public test. Research: Revise the CDN paper 21 Description (02/23/2004-02/29/2004) I have finished 2/3 of the rapid spanning tree protocol (RSTP) test cases. The primarily test result shows there are less failures than the spanning tree protocol (RTP). I think the actual reason of less failures could be found when the final test result come out. Testing RSTP is much harder than testing general RTP because the RSTP involves more transition states, and more complex mechanisms and algorithms. The multiple spanning tree protocol (MSTP) is the most complex; hence its test suite has 231 test cases. To perform the testing software is easy, but to understand the standards and to find out failure reasons are the hardest things. 22 Works Done Last Week (02/23/2004-02/29/2004) NBL: Intern student: finish the STP conformance testing Accton: perform the spanning tree protocols conformance testing (RSTP, MSTP) Done, but the price is still too high. So I decide to buy the devices from the Internet. Discuss the testing program for the ADSL public test. The pass ratio of RSTP is 82.7%. Ask for a new quotation of the PDU and terminal server Done. The pass ratio of Cisco 2950 is 88.7%. Done. Research: Revise the CDN paper Still on going. 23 Works Planned for This Week (03/01/2004-03/07/2004) NBL: Intern student: perform the RSTP conformance testing Accton: perform the VLAN conformance testing Buy the PDU and terminal server from the Internet Finish the final version of the bridging and routing test items Response the request of testing PPPoE from Airocon Research: Revise the CDN paper 24 Description (03/01/2004-03/07/2004) The cost for constructing a complete testbed is high. With enough budget, everything can be done easier. If not, we have to come out a cost-effective solution. The reason we choose Ixia solutions is that we can do more layer 4-7 tests than using Spirent solutions. However, if the budget is really limited, we have to re-think what devices to buy. I think one or two Ixia modules (8 ports or 16 ports) is enough for most cases. I will investigate the requests from the other PMs as soon as possible. 25 Works Done Last Week (03/01/2004-03/07/2004) NBL: Intern student: perform the RSTP conformance testing Accton: perform the VLAN conformance testing Done. Response the request of testing PPPoE from Airocon I have ordered the devices from TechOnWeb website Finish the final version of the bridging and routing test items Half done. Buy the PDU and terminal server from the Internet Done. The pass ratio of Cisco Catalyst 2950 is 70%. Done. Airocon asked for a quotation for the PPPoE conformance testing. Research: Revise the CDN paper Still on going. 26 Works Planned for This Week (03/08/2004-03/14/2004) NBL: Accton: Start to test RIP conformance Wait for the final order confirmation from TechOnWeb Prepare to test PPPoE of Airocon SOHO router Research: Revise the CDN paper 27 Description (03/08/2004-03/14/2004) There are some drawbacks to use WebBench to measure Web performance. We need lots of client computers to generate enough traffic for a Web server. But the cost of buying computer and the replacement rate are high. Spirent’s Avalanche and Reflector can emulate Web clients and servers precisely and the scalability is enough to test high-performance Web server, even Web switch. But the same problem is the cost of buying Spirent solutions is hight. 28 Works Done Last Week (03/08/2004-03/14/2004) NBL: Accton: Start to test RIP conformance Wait for the final order confirmation from TechOnWeb Done. The total cost is $1974,93. I have asked Su-Yu to follow up the order. Prepare to test PPPoE of Airocon SOHO router Done. The pass ratio is 96%. Airocon still did not give any response. I have read PPP related RFCs. Research: Revise the CDN paper Still on going. 29 Works Planned for This Week (03/15/2004-03/21/2004) NBL: Accton: Start to test OSPF conformance Setup a PPPoe testing environment Research: Revise Shuo-Wen’s outline and Chapter 1. Revise the CDN paper 30 Description (03/15/2004-03/21/2004) The Point-to-Point Protocol (PPP) provides a standard method for transporting packets over point-to-point link. PPP over Ethernet (PPPoE) allows PPP transmission over Ethernet. It combines the advantages of the low cost Ethernet media and the advantages of dial-up connection. Using PPPoE, a service provide can allocate IP address dynamically to home users and track a specific username and password. The provider can also distinguish between different subscribers and bill them accordingly. More and more home users choose PPPoE through ADSL for Internet connection and discard the traditional dial-up through analog modem. 31 Works Done Last Week (03/15/2004-03/21/2004) NBL: Accton: Start to test OSPF conformance Setup a PPPoE testing environment Still on going The PPPoE server was constructed. Research: Revise Shuo-Wen’s outline and Chapter 1. Done Revise the CDN paper Still on going 32 Works Planned for This Week (03/22/2004-03/28/2004) NBL: Accton: Continue to test OSPF conformance Make slides of bridging and routing testing services Write a experiment manual for bridging testing Research: Revise Shuo-Wen’s Chapter 2. Prepare for the proposal presentation in the next week Revise the CDN paper 33 Description (03/22/2004-03/28/2004) There are some differences between RIPv1 (Routing Information Protocol version 1) and RIPv2. First, RIPv1 does not support Variable Length Subnet Mask (VLSM) but RIPv2 does. This means RIPv1 is classful and RIPv2 is classless. Next, RIPv2 supports authentication of routing updates. The authentication can be a simple password or a message digest (e.g. MD5). Next, RIPv2 supports multicast route updates, instead of broadcasting update messages to all connected neighbor routers in RIPv1. Finally, RIPv2 carries next hop address with each route entry, but RIPv1 does not. 34 Works Done Last Week (03/22/2004-03/28/2004) NBL: Accton: Continue to test OSPF conformance Make slides of bridging and routing testing services Done. Write a experiment manual for bridging testing 40% of the test cases were done. The manual and slides were written. Research: Revise Shuo-Wen’s Chapter 2. Prepare for the proposal presentation in the next week Not yet. Done. Revise the CDN paper Still on going. 35 Works Planned for This Week (03/29/2004-04/04/2004) NBL: Accton: Continue to test OSPF conformance Move bridging and routing testbed from 701 to 604 Add STP interoperability testing to the experiment manual Research: Revise Shuo-Wen’s Chapter 2. Revise the CDN paper 36 Description (03/29/2004-04/04/2004) RIP has some limitations that could cause problems in large networks. First, RIP has a limit of 15 hops. A RIP network that spans more than 15 hops is considered unreachable. Second, RIP cannot handle variable length subnet masks, which means RIP is a classful routing protocol. Third, periodic broadcasts of the full routing table consume a large amount of bandwidth. This is a major problem with large networks especially on slow links. Finally, RIP networks are flat networks. There is no concepts of areas or boundaries. Thus, RIP is not suitable for large networks. 37 Works Done Last Week (03/29/2004-04/04/2004) NBL: Accton: Continue to test OSPF conformance Move bridging and routing testbed from 701 to 604 Done. Bridging and routing marketing The racks have been moved to 604. Add STP interoperability testing to the experiment manual About 200 test cases were done. I have sent the service documents to 11 vendors. ICPlus asked for a presentation on 4/16 Research: Revise Shuo-Wen’s Chapter 2. Done. Revise the CDN paper Still on going. 38 Works Planned for This Week (04/05/2004-04/11/2004) NBL: Accton: Continue to test OSPF conformance Move bridging and routing testbed from 701 to 604 Track the response of the marketing actions Research: Revise Shuo-Wen’s Chapter 3. Prepare for the mid-term exam 39 Description (04/05/2004-04/11/2004) There are several scheduling disciplines such as Priority Queuing (PQ), Weighted Fair Queuing (WFQ), Weighted Round Robin (WRR), and Deficit Round Robin (DRR) can be used for determining which request to be fetched from a class queue. In PQ, requests are scheduled from the head of a given queue only if all queues of high priority are empty. Within each of the priority queues, requests are scheduled in First In First Out (FIFO) order. However, if the volume of higher-priority requests becomes excessive, lower-priority requests can be dropped as the buffer space allocated to low-priority queues start to overflow. Beside, the head-of-line blocking of FIFO queue makes the requests behind the head be delayed or even be dropped due to buffer overflow. 40 Works Done Last Week (04/05/2004-04/11/2004) NBL: Accton: Continue to test OSPF conformance Move bridging and routing testbed from 701 to 604 Done. Track the response of the marketing actions Done. The pass ratio is 86%. ICPlus and Delta Networks asked for a presentation to them. Research: Revise Shuo-Wen’s Chapter 3. Not yet. Prepare for the mid-term exam On going. 41 Works Planned for This Week (04/12/2004-04/18/2004) NBL: Accton: Begin to test OSPF extensions conformance Visit ICPlus and Delta Networks Research: Revise Shuo-Wen’s Chapter 3. Prepare for the mid-term exam 42 Description (04/12/2004-04/18/2004) Given different classes of clients and a web server, our goal is to provide service differentiation through HTTP request scheduling on the website gateway. All HTTP requests originated from clients will pass through the website gateway and be scheduled to the web server according to the QoS policies and the availability of the server resources. The website gateway first classifies the incoming requests into different service classes by inspecting the contents of IP and HTTP headers. Then the scheduler decides which request can be fetched from the class queues and when to release it to the web server according to the QoS policies and the availability of the server resources, respectively. 43 Works Done Last Week (04/19/2004-04/25/2004) NBL: Accton: Begin to test IGMP conformance Interview the new part-time candidates Delta Networks is arranging. Loops asked for testing a WAN router. Present the bridging experiment in the Network Experiment class Done. Arrange the visits to the vendors and track responses Not finished yet. Done. Research: Revise Shuo-Wen’s Chapter 4 Not yet. Revise the CDN paper Not finished yet. 44 Works Planned for This Week (04/26/2004-05/02/2004) NBL: Accton: Begin to test DVMRP conformance Train the new parttime student Arrange the visits to the vendors and track responses Prepare for the Loop WAN router testing Prepare for the Alcatel switch testing Research: Revise Shuo-Wen’s Chapter 4 Revise the CDN paper 45 Description (04/26/2004-05/02/2004) The speed of the modified DRR scheduling should be throttled such that the released request would not overwhelm the server capacity. For this, the token bucket mechanism is employed to decide when to release a request to the server. The current utilization of a server resource is represented by a token bucket. The size of a token bucket, i.e. server resource capacity, and the amounts of resources (tokens) a request needs is estimated from the results of the server probing. When a request is released, the token buckets discard tokens. Conversely, when an HTTP response finishes passing through the gateway, i.e. the server resources are released, the discarded tokens are returned the buckets. 46 Works Done Last Week (04/26/2004-05/02/2004) NBL: Accton: Begin to test DVMRP conformance Train the new parttime student Loop had set two WAN routers to NBL. Prepare for the Alcatel switch testing No new response. Prepare for the Loop WAN router testing I asked him to read some training materials and prepare for the layer 2 switch testing. Arrange the visits to the vendors and track responses Done. The pass ratio is 47.0% Postponed to this week. Research: Revise Shuo-Wen’s Chapter 4 Not yet. Instead, I revised the Chapter 1 to 3 and the slides again. Revise the CDN paper Not finished yet. 47 Works Planned for This Week (05/03/2004-05/09/2004) NBL: Accton: Begin to test PM-DM conformance Discuss the WAN router testing with Loop Prepare for the Alcatel switch testing Research: Revise Shuo-Wen’s Chapter 4 Revise the CDN paper 48 Description (05/03/2004-05/09/2004) For probing the characteristics of the web pages, the server prober first retrieves the homepage of the website, parses the homepage to find embedded object and other hyperlinks. The probers recursively scans the web pages on the same sever link by link until all the web pages have be scanned. The probed URL, response size and the amounts of required resources of each page will be recorded in the web page table. Note that the response size of a page recorded in the table is chosen from a number close to the actual probed responses size. 49 Works Done Last Week (05/03/2004-05/09/2004) NBL: Accton: Begin to test PM-DM conformance Discuss the WAN router testing with Loop Done. Prepare for the Alcatel switch testing On-going. The test will be finished this week. The VLAN functionality was validated. Research: Revise Shuo-Wen’s Chapter 4 Not finished yet. Revise the CDN paper Not finished yet. 50 Works Planned for This Week (05/10/2004-05/16/2004) NBL: Accton: Finish the PIM-DM conformance testing and begin to review the log files Write a test plan for the Loop WAN router testing Buy 3 PCs for the testing environment Research: Revise Shuo-Wen’s Chapter 4 Revise the CDN paper 51 Description (05/10/2004-05/16/2004) In typical DRR, the speed of the scheduling is expected as fast as possible so as to maximize the performance. However, the server capacity is finite, it cannot processes unlimited number of request at a time, or the user-perceived latency would be longer. So the scheduler should release requests according to the availability of the server resources. In our work, we employ token bucket mechanism to throttle the speed of releasing requests. Each token bucket stands for the current utilization of a server resource. When a request is released, the token buckets discard tokens. On the contrary, When an HTTP response finishes passing through the gateway, i.e. the server resources are released, the discarded tokens are returned the buckets. 52 Works Done Last Week (05/10/2004-05/16/2004) NBL: Accton: Finish the PIM-DM conformance testing and begin to review the log files Write a test plan for the Loop WAN router testing On going. The test plan will be ready by this Thursday. Buy 3 PCs for the testing environment The pass ratio of the PIM-DM conformance testing is 30.4%. Done. The testing software was installed. Research: Revise Shuo-Wen’s Chapter 4 On-going. Revise the CDN paper On-going. 53 Works Planned for This Week (05/17/2004-05/23/2004) NBL: Accton: Finish the DHCP conformance testing and begin to review the log files Write a test plan for the Loop WAN router testing Research: Revise Shuo-Wen’s Chapter 4 Revise the CDN paper 54 Description (05/17/2004-05/23/2004) The purpose of the request scheduler is to fetch requests from the class queues and to schedule the fetched request to the web server. In other words, the request scheduler decides which request can be fetched next from the class queues and when to release it to the server. The deficit round robin scheduling is employed as the scheduling algorithm to decide which request can be fetched next, that is, to partition the server resource proportionally according to the quanta defined in the QoS policy table. On the other hand, the sliding window mechanism is employed to throttle the speed of releasing requests so as not to overwhelm the server capacity. 55 Works Done Last Week (05/17/2004-05/23/2004) NBL: Accton: Finish the DHCP conformance testing and begin to review the log files Write a test plan for the Loop WAN router testing Done. The TCP performance and the STP conformance tests are added. Research: Revise Shuo-Wen’s Chapter 4 All done. Revise the CDN paper Not finished yet. 56 Works Planned for This Week (05/24/2004-05/30/2004) NBL: Accton: Review the test logs to generate bug lists. IP forwarding, NATP and DHCP functionality testing for the Loop WAN router. Research: Revise the CDN paper 57 Description (05/24/2004-05/30/2004) The effect of the services differentiation can be evaluated on the throughput and the userperceived latency. The aggregated throughput and the user-perceived latency of a service class are measured for a comparison between the activation and the deactivation of the request scheduling. The measurement is performed with fixed-sized and mixed-sized Web pages to demonstrate the stability would remain regardless the size of the Web pages. 58 Works Done Last Week (05/24/2004-05/30/2004) NBL: Accton: Review the test logs to generate bug lists. Loop WAN router: IP forwarding, NATP and DHCP functionality testing The IP Gateway, RIP, and partially OSPF_CORE logs have been reviewed. Done. The pass ratio is 100%. Research: Revise the CDN paper Not finished yet 59 Works Planned for This Week (05/31/2004-06/06/2004) NBL: Accton: Loop WAN router: Write a mid-term report containing bug NAPT, Virtual Service and SNTP functionality testing Research: Revise the CDN paper 60 Description (05/31/2004-06/06/2004) The request classifier and the request scheduler are implemented as a daemon program called “WebQ” on the NetBSD system. The program runs at the user space and listens on the port 880 of the loopback IP address, i.e. 127.0.0.1:880. To make the WebQ work transparent to both clients and the Web server, the ipnat rewrites the destination IP address and the port number of the incoming HTTP packets to redirect the request to the WebQ for service differentiation. The WebQ performs the request classification and the request scheduling and sends the request to the Web server. The HTTP responses from the Web server also pass through the WebQ and return to the clients. 61 Works Done Last Week (05/31/2004-06/06/2004) NBL: Accton: Write a mid-term report containing bug Loop WAN router: NAPT, Virtual Service and SNTP functionality testing Partially done. Done. The pass ratios are 88.9%, 50%, and 100%, respectively. Research: Revise the CDN paper Not finished yet. 62 Works Planned for This Week (06/07/2004-06/13/2004) NBL: Accton: Loop WAN router: Finish the mid-term report. TCP, HTTP, and RIP performance testing Research: Revise the CDN paper 63 Description (06/07/2004-06/13/2004) The evaluation environment consists of an Apache Web server, the WebQ gateway, and several computers running the WebBench Web performance testing tool. The work load files of the WebBench are put to the Web server prior to the measurement. The WebBench controller orders the WebBench clients to issue HTTP requests to the Web server and gathers the test data from the WebBench clients. In the service differentiation evaluation, the WebBench clients are divided into three service classes, whose ratio of the service weights are set to 6:3:1. 64 Works Done Last Week (06/07/2004-06/13/2004) NBL: Accton: Finish the mid-term report. Loop WAN router: TCP, HTTP, and RIP performance testing Done. Done Research: Revise the CDN paper Not finished yet. 65 Works Planned for This Week (06/14/2004-06/20/2004) NBL: Accton: Loop WAN router: Start to design feature combination test cases. STP and IP Gateway conformance testing Research: Prepare for the formal language final exam. Revise the CDN paper 66 Description (06/14/2004-06/20/2004) The effect of the services differentiation can be evaluated on the throughput and the userperceived latency. The aggregated throughput and the user-perceived latency of a service class are measured for a comparison between the activation and the deactivation of the request scheduling. The measurement is performed with fixed-sized and mixed-sized Web pages to demonstrate the stability would remain regardless the size of the Web pages. 67 Works Done Last Week (06/14/2004-06/20/2004) NBL: Accton: Start to design feature combination test cases. Loop WAN router: STP and IP Gateway conformance testing On going. Done. Research: Prepare for the formal language final exam. Done. Revise the CDN paper Not finished yet. 68 Works Planned for This Week (06/21/2004-06/27/2004) NBL: Accton: Loop WAN router: Design feature combination test cases. Help Loop to debug RIP crash problems Research: Prepare the network experiment final exam. Revise the CDN paper 69 Description (06/21/2004-06/27/2004) In content-centric CDN and access-centric CDN, hosting providers do not seem to profit too much from providing collocation services to CDN service providers. This is because content providers may deal with CDN service providers instead of hosting providers to distribute their contents. As an example, five original customers of a hosting provider turn to employ distribution service of a CDN service provider, one of whose edge servers is collocated in the datacenter of this hosting provider. Therefore, the two business models may be not suitable for hosting providers. Furthermore, the scalability of contentcentric CDN and access-centric CDN is limited because a CDN service provider cannot expand its network infinitely. 70 Works Done Last Week (06/21/2004-06/27/2004) NBL: Accton: Design feature combination test cases. Loop WAN router: Help Loop to debug RIP crash problems On going. Done. Research: Prepare the network experiment final exam. Done. Revise the CDN paper Still on going. 71 Works Planned for This Week (06/28/2004-07/04/2004) NBL: Accton: Loop WAN router: Design feature combination test cases. Finish all the tests and send out the test results Survey switch and router vendors and write marketing log Research: Revise the CDN paper 72 Description (06/28/2004-07/04/2004) Internet content delivery has evolved from centralized hosting, Web replication to recent content delivery network (CDN). Centralized hosting has the problems of high backbone bandwidth consumption and long user-perceived latency, while Web replication leads to high investment and operational costs in deploying servers in several networks. However, content delivery networks promise to answer these problems. In this paper, we introduce the evolution of Internet content delivery and CDN business models, and evaluate the performance of CDN. 73 Works Done Last Week (06/28/2004-07/04/2004) NBL: Accton: Design feature combination test cases. Loop WAN router: Finish all the tests and send out the test results Done, except the filtering performance. Survey switch and router vendors and write marketing log The RIP combination is done. Done. Research: Revise the CDN paper Done. 74 Works Planned for This Week (07/05/2004-07/11/2004) NBL: Accton: Loop WAN router: Mid-term report. Design feature combination test cases. Finish all the tests and send out the test results Find correct persons of switch and router vendors Research: Format the CDN paper and submit. 75 Description (07/05/2004-07/11/2004) Because most end users set their browsers to use caching proxies and ISPs deploy transparent proxies to intercept requests of end users, end users first connect to caching proxies and then to origin servers only if cache miss occurs. Some content providers may wish to collect access logs of servers for knowing how many users browse their sites. Without the logs of caching proxies, the statistical results will be incorrect. However, it is not easy to collect all the logs for content providers because caching proxies may not belong to only one ISP. 76 Works Done Last Week (07/05/2004-07/11/2004) NBL: Accton: Mid-term report. Design feature combination test cases. Finish all the tests and send out the test results Done. Find correct persons of switch and router vendors The routing protocol combination was done. Loop WAN router: Done The introduction documents were sent to Ambit and Asus. Research: Format the CDN paper and submit. On going. 77 Works Planned for This Week (07/12/2004-07/18/2004) NBL: Accton: Loop WAN router: Design feature combination test cases of multicast routing protocols Write the final test report. Find correct persons of switch and router vendors Research: Format the CDN paper and submit. 78 Description (07/12/2004-07/18/2004) Internet backbones are composed of several long-haul networks. As the growth of Internet traffic, the capacity of backbones must meet all requirements, otherwise insufficient capacity of backbones will lead to backbone problems. Generally, the capacity of a network is decided by links and routers. To solve backbone problems, one can upgrade the capacities of links and routers or reduce traffic flowing through backbones. An approach to reducing traffic is that ISPs build forward caching proxies close to end users, storing copies of downloaded files from origin servers. By intercepting requests of end users, a forward caching proxy returns cached contents to end users directly. 79 Works Done Last Week (07/12/2004-07/18/2004) NBL: Accton: Design feature combination test cases of multicast routing protocols Loop WAN router: Write the final test report. On going Find correct persons of switch and router vendors done The visit to Delta Networks has been arranged. Research: Format the CDN paper and submit. On-going. 80 Works Planned for This Week (07/19/2004-07/25/2004) NBL: Accton: Loop WAN router: Finish the final test report. Marketing: Design feature combination test cases of layer 2 protocols Visit Delta Networks on 7/22 Find correct persons of switch and router vendors Research: Format the CDN paper and submit. 81 Description (07/19/2004-07/25/2004) A peering point is the connection between two Internet service providers. Peering allows very fast and highquality connectivity between the two partners. Peering problems generally result from business concerns, not performance problems. As an example, if there are two ISPs investing a large amount of money in constructing their network infrastructures, they might be unwilling to forward packets for each other for the sake of competition. This may make the packets from one ISP have to traverse a suboptimal path and thus decrease the speed of content delivery. One way to solving the peering problem is that ISPs sign peering agreements and forward packets for each other. 82 Works Done Last Week (07/19/2004-07/25/2004) NBL: Accton: Design feature combination test cases of layer 2 protocols Loop WAN router: Finish the final test report. Done Marketing: Visit Delta Networks on 7/22 Done Find correct persons of switch and router vendors done On going Research: Format the CDN paper and submit. No finished yet. 83 Works Planned for This Week (07/26/2004-08/01/2004) NBL: Accton: Loop WAN router: Send the final test report. Make slides Marketing: Design feature combination test cases of layer 3 protocols Track the marketing status of the switch and router vendors Research: Format the CDN paper and submit. 84 Description (07/26/2004-08/01/2004) Content delivery network (CDN) is developed to solve these bottlenecks. CDSPs build edge servers close to end users in one or more ISP networks and actively push contents to the edge servers for ICPs; thus requests of end users are serviced by the edge servers directly. Because the pushed contents on the edge servers can be accessed by a large number of requests from end users, load on origin servers, bandwidth consumption of backbones, and user-perceived latency can be reduced effectively. Furthermore, content providers delegate contents to only one CDN service provider instead of multiple ISPs or hosting providers. The cost of the infrastructure is shared by all CDN participants. 85 Works Done Last Week (07/26/2004-08/01/2004) NBL: Accton: Design feature combination test cases of layer 3 protocols Loop WAN router: Send the final test report. On-going Marketing: Track the marketing status of the switch and router vendors Done. Make slides On-going. On-going. Research: Format the CDN paper and submit. Done. 86 Works Planned for This Week (08/02/2004-08/08/2004) NBL: Accton: Loop WAN router: Make slides Marketing: Design feature combination test cases of layer 3 protocols Track the marketing status of the switch and router vendors Research: 87 Description (08/02/2004-08/08/2004) The developments of content delivery networks are toward more efficient and divers delivery schemes. For example, caching dynamic contents on edge servers can reduce bandwidth consumption and user-perceived latency. Optimizing delivery of streaming media to bypass hotspots can provide better quality of service for end users. Moreover, there is a trend of providing personalization services to end users. As an example, Amazon bookstore analyzes the query logs and predicts the preferences of end users for recommending new books to them. In addition, executing services on edge servers instead of origin servers can offload origin servers and reduce latency. In a word, pushing applications to edge servers or providing value-added services on edge server are future trends. 88 Works Done Last Week (08/02/2004-08/08/2004) NBL: Accton: Design feature combination test cases of layer 3 protocols Loop WAN router: Make slides Done. The date for the report is 8/11. Marketing: Track the marketing status of the switch and router vendors The configuration figures are ready. On going. Research: 89 Works Planned for This Week (08/09/2004-08/15/2004) NBL: Accton: Loop WAN router: Make a presentation on 8/11. Marketing: Finish the final report and the slides. Track the marketing status of the switch and router vendors Research: 90 Description (08/09/2004-08/15/2004) Internet content delivery has evolved from centralized hosting, Web replication to the recent content delivery network (CDN). Centralized hosting has the problems of high backbone bandwidth consumption and long user-perceived latency, while Web replication leads to high investment and operational costs in deploying servers in several ISP networks. However, a content delivery network acts as a third party to deliver content for various content providers into ISPs networks. In this paper, we introduce the evolution of Internet content delivery and evaluate the performance of CDN. In our evaluation, we find that CDN saves more bandwidth consumption and distribution time than traditional distribution alternatives, has shorter user-perceived latency than centralized hosting, and improves the delivery quality of streaming media. 91 Works Done Last Week (08/09/2004-08/15/2004) NBL: Accton: Finish the final report and the slides. Loop WAN router: Make a presentation on 8/11. Done. Marketing: Track the marketing status of the switch and router vendors Done. Paused in this week. Research: 92 Works Planned for This Week (08/16/2004-08/22/2004) NBL: Accton: NBL testing technology workshop: Make slides Marketing: Write a follow-up cooperation proposal Start to write a test plan for performance testing. Track the marketing status of the switch and router vendors Research: 93 Description (08/16/2004-08/22/2004) The possible forms of the follow-up cooperation are summarized as follows: 1. Start-up: Accton invests in establishing a start-up for network testing. NBL people operate the start-up. But the time is not good for NBL to change into a company nor to establish a startup-up now. 2. Donation: Accton donates money to NBL, whereas NBL comments to offer human resources and test expertises to assist Accton in testing. The wished amount of money is the double of the execution cost. That is, a half is for the execution cost, the other half is indeed the donation. 3. Half-year Project: NBL people work for an Accton project. The goal of the cooperation project is mentioned above. Accton pays 500K to NBL and pays 6-month salary to the people who work at Accton. 94 Works Done Last Week (08/16/2004-08/22/2004) NBL: Accton: Write a follow-up cooperation proposal Start to write a test plan for performance testing. On-going NBL testing technology workshop: Make slides Done done Research: 95 Works Planned for This Week (08/23/2004-08/29/2004) NBL: Accton: Start to construct a system testbed Start to write a test plan for performance testing NBL testing technology workshop Research: Revise Shou-Yen’s patent 96 Description (08/23/2004-08/29/2004) Power-over-Ethernet (PoE) eliminates the need to run 110/220V power to wireless accesspoints and other devices on a wired LAN. Using PoE, system installers need to run only a single CAT5 Ethernet cable that carries both power and data to each device. This allows greater flexibility in the locating of AP's and network devices and significantly decreasing installation costs in many cases. 97 Works Done Last Week (08/23/2004-08/29/2004) NBL: Accton: Start to construct a system testbed Start to write a test plan for performance testing On-going. NBL testing technology workshop The devices are ready and the firmware is upgraded. done Research: Revise Shou-Yen’s patent On-going 98 Works Planned for This Week (08/30/2004-09/05/2004) NBL: Accton: Workshop follow-ups: Connect devices and configure software Write a test plan for performance testing Write a half-year cooperation proposal Send test items and deal with possible new cases Research: Revise Shou-Yen’s patent 99 Description (08/30/2004-09/05/2004) Ethernet, Fast Ethernet and Gigabit Ethernet have clearly been adopted as the technologies of choice for building high-performance local area networks. 10 Gigabit Ethernet is simply the next development in this Ethernet bandwidth hierarchy. An evolutionary step forward, 10 Gigabit Ethernet will preserve many of the same characteristics of previous versions of Ethernet, continuing the Ethernet tradition of providing the most cost-effective and simple-to-implement networking technology. 100 Works Done Last Week (08/30/2004-09/05/2004) NBL: Accton: Connect devices and configure software Write a test plan for performance testing Done. The proposal and the contract has been sent to Accton Workshop follow-ups: Send test items and deal with possible new cases suspended Write a half-year cooperation proposal done done Research: Revise Shou-Yen’s patent On-going 101 Works Planned for This Week (09/06/2004-09/12/2004) NBL: Accton: Workshop follow-ups: Perform the system test test plan Wait for response from Realtek Appoint a time to visit VIA networks Research: Revise Shou-Yen’s patent 102 Description (09/06/2004-09/12/2004) Power-over-Ethernet (PoE) eliminates the need to run 110/220V power to wireless accesspoints and other devices on a wired LAN. Using PoE, system installers need to run only a single CAT5 Ethernet cable that carries both power and data to each device. This allows greater flexibility in the locating of AP's and network devices and significantly decreasing installation costs in many cases. 103 Works Done Last Week (09/06/2004-09/12/2004) NBL: Accton: Perform the system test test plan Workshop follow-ups: Wait for response from Realtek No new resoponse Appoint a time to visit VIA networks 1st round finished. The visit is on 9/15 afternoon Research: Revise Shou-Yen’s patent Done. 104 Works Planned for This Week (09/13/2004-09/19/2004) NBL: Accton: Workshop follow-ups: Revise the system test test cases Visit VIA networks on 9/15 Send NBL column article to Connectimes Ask for the Asus review article Research: Send Shou-Yen’s patent to ITRI 105 Description (09/13/2004-09/19/2004) Once the scheduler fetches a request, the next step is to decide when to release it to the Web server. First assume the request-sending rate of the clients is higher than the request-processing rate of the server. If the rate of the scheduling is unbounded, the incoming requests will be sent to the server without being queued on the gateway. This may make the requests be queued on the server and even overwhelm the server. 106 Works Done Last Week (09/13/2004-09/19/2004) NBL: Accton: Revise the system test test cases Workshop follow-ups: Visit VIA networks on 9/15 Done. Ask for the Asus review article Done. Send NBL column article to Connectimes Add “ARP Cache Capacity” test case Asus decided to send WLAN AP for the review Research: Send Shou-Yen’s patent to ITRI Done. 107 Works Planned for This Week (09/20/2004-09/26/2004) NBL: Accton: Add new test cases to system test Deal with the possible case from Edimax Research: Start to revise Shou-Yen’s paper 108 Description (09/20/2004-09/26/2004) The effect of the services differentiation can be evaluated on the throughput and the userperceived latency. The aggregated throughput and the user-perceived latency of a service class are measured for a comparison between the activation and the deactivation of the request scheduling. The measurement is performed with fixed-sized and mixed-sized Web pages to demonstrate the stability would remain regardless the size of the Web pages. 109 Works Done Last Week (09/20/2004-09/26/2004) NBL: Accton: Add new test cases to system test Deal with the possible case from Edimax Hawking Networks asked for a conference call on 9/30 Track the response from VIA Networking Done. VIA Networking asked for a quote. Research: Start to revise Shou-Yen’s paper On-going. Revise aging Shou-Yen’s patent Done. 110 Works Planned for This Week (09/27/2004-10/03/2004) NBL: Accton: Add new test cases to system test Track the half-year cooperation Conference call with Hawking Networks Send a quote to VIA Networking Research: Revise Shou-Yen’s paper 111 Description (09/27/2004-10/03/2004) The amount of disk I/O or bandwidth needed for a specific page-size request is easy to estimate. The value from dividing the throughput by the processing rate is the answer. However, the processing power a request needs cannot be derived directly from the above simple calculation. Assume that the CPU usage is the same among the requests of the same page size and the CPU usage of the non-CPU-intensive requests is very little or can be even ignored. 112 Works Done Last Week (09/27/2004-10/03/2004) NBL: Accton: Add new test cases to system test Track the half-year cooperation Done. Send a quote to VIA Networking The price and the jobs are discussed. Conference call with Hawking Networks The “Topology Change” test case is modified. Done. Research: Revise Shou-Yen’s paper On-going. 113 Works Planned for This Week (10/04/2004-10/10/2004) NBL: Accton: Add new test cases to system test Modify the half-year cooperation contact and the proposal Read resumes and interview job applicants Host Prof. Huang from NTU EE Research: Revise Shou-Yen’s paper 114 Description (10/04/2004-10/10/2004) The server prober has to probe the resource capacities of the server and the resources requirement (CPU and I/O usage) of Web pages. For knowing the resource capacities of the server, the testing methodology of last example can be employed. First, test the maximum request processing rate and the maximum throughput under different page size. Then the upper bound of request processing rate and throughput, which represent the CPU and I/O capacity respectively, can be known. 115 Works Done Last Week (10/04/2004-10/10/2004) NBL: Accton: Add new test cases to system test Modify the half-year cooperation contact and the proposal An interview is appointed on 10/12. Host Prof. Huang from NTU EE Done. Read resumes and interview job applicants HSRP and VRRP are modified. ARP and static routing stress tests are added. Done. Research: Revise Shou-Yen’s paper On-going. 116 Works Planned for This Week (10/11/2004-10/17/2004) NBL: Accton: Add new test cases to system test Track the status of the contract Interview a job applicant Track the possible case from VIA Networking Plan SIG and 3-day training course Research: Revise Shou-Yen’s paper 117 Description (10/11/2004-10/17/2004) In order to know the resource capacities of the server (e.g. processing power, disk I/O speed, and network bandwidth) and the characteristics of the Web pages (response size and resource requirement), the server prober probes the server capacity and Web pages before on-line operation. First, which kinds of the server resource required by the Web pages should be identified. 118 Works Done Last Week (10/11/2004-10/17/2004) NBL: Accton: Add new test cases to system test Track the status of the contract On-going Plan SIG and 3-day training course On-going. Track the possible case from VIA Networking The final version is passed to Accton. Interview a job applicant “VLAN scalability” and “DHCP server scalability” tests are added. Done. Research: Revise Shou-Yen’s paper On-going 119 Works Planned for This Week (10/18/2004-10/24/2004) NBL: Accton: Add new test cases to system test Track the status of the contract Review resumes and interview job applicants Research: Revise Shou-Yen’s paper 120 Description (10/18/2004-10/24/2004) From the observations, the server prober has to probe the resource capacities of the server and the resources requirement (CPU and I/O usage) of Web pages. For knowing the resource capacities of the server, a testing methodology can be employed. First, test the maximum request processing rate and the maximum throughput under different page size. Then the upper bound of request processing rate and throughput, which represent the CPU and I/O capacity respectively, can be known. 121 Works Done Last Week (10/18/2004-10/24/2004) NBL: Accton: Add new test cases to system test Track the status of the contract Cooper is dealing with this. Review resumes and interview job applicants Done On-going. Research: Revise Shou-Yen’s paper Done. 122 Works Planned for This Week (10/25/2004-10/31/2004) NBL: Accton: Finish the feature combination test Write the feature combination test report Track the status of the contract Review resumes and interview job applicants Research: Submit the Web QoS paper to Internet Computing 123 Description (10/25/2004-10/31/2004) With the explosive growth of Web traffic, the load on a Web server becomes heavier, leading to the longer userperceived latency. Website operators would like to employ service differentiation to offer better throughput and user-perceived latency to some specific users. This paper presents an HTTP request scheduling algorithm deployed at the website gateway to enable the Web quality of service without any modification to client or server software. A variation of the deficit round robin scheduling algorithm and a window control mechanism are presented to decide the order and the release time of requests, respectively. 124 Works Done Last Week (10/25/2004-10/31/2004) NBL: Accton: Finish the feature combination test Write the feature combination test report On-going. Review resumes and interview job applicants Done. Track the status of the contract Done. On-going. Research: Submit the Web QoS paper to Internet Computing The paper was reviewed by yclai. 125 Works Planned for This Week (11/01/2004-11/07/2004) NBL: Accton: Write the switch performance test plan Review resumes and interview job applicants Write a paragraph for the NBL newsletter Research: Revise Web QoS paper according to yclai’s comment Submit the Web QoS paper to Internet Computing 126 Description (11/01/2004-11/07/2004) To reduce the user-perceived latency, the bottleneck of accessing a Web page should be identified first. An example of user-perceived latency measurement on downloading the homepages of 40 important US-based business websites from Keynote in 2001 is given here. Keynote company measured website performance from its 1,700 measurement computers distributed in 50 metropolitan areas worldwide The user-perceived latency, i.e. Web page download time, is decomposed into six components: 1) DNS (Domain Name System) lookup time, 2) TCP (Transmission Control Protocol) connection time, 3) server delay time, 4) redirection time, 5) base page download time, and 6) content download time. 127 Works Done Last Week (11/01/2004-11/07/2004) NBL: Accton: Write the switch performance test plan Review resumes and interview job applicants One applicant was interviewed, but no qualified. Write a paragraph for the NBL newsletter On-going Done. Research: Revise Web QoS paper according to yclai’s comment Done. Submit the Web QoS paper to Internet Computing Done. 128 Works Planned for This Week (11/08/2004-11/14/2004) NBL: Write the switch performance test plan for Accton Review resumes and interview job applicants Write training materials for the bridging and routing testing Research: Find the suitable journals for the submission of the CDN paper Think about the multi-resource scheduling algorithm for Web clusters 129 Description (11/08/2004-11/14/2004) We focus on resolving server bottleneck because website operators can completely control their servers, but cannot do much on improving the whole network performance. The goals are to improve the server throughput and reduce the user-perceived latency for some specific users, in other words, to provide service differentiation at the server side, and thus allows some users to perceive the shorter latency on downloading Web pages. 130 Works Done Last Week (11/08/2004-11/14/2004) NBL: Write the switch performance test plan for Accton Review resumes and interview job applicants Writing RFC 2889 testing Write the SIG plan One applicant is interviewed, but not qualified Write training materials for the bridging and routing testing Write RFC 2889 testing Done. Research: Find the suitable journals for the submission of the CDN paper On-going. Think about the multi-resource scheduling algorithm for Web clusters On-going. 131 Works Planned for This Week (11/15/2004-11/21/2004) NBL: Write the switch performance test plan for Accton Review resumes and interview job applicants Write training materials for the bridging and routing testing Research: Find the suitable journals for the submission of the CDN paper Think about the multi-resource scheduling algorithm for Web clusters 132 Description (11/15/2004-11/21/2004) RIP is a routing protocol based on the Bellman-Ford (or distance vector) algorithm. In a RIP network, each RIP router sends RIP update messages to its neighbor routers to exchange routing information. Each RIP router keeps a RIP routing table listing the best route to every destination network. OSPF is routing protocol based on the Dijkstra’s Shortest Path First (SPF) algorithm. OSPF creates and maintains a distributed replicated database (link-state database) that contains a complete picture of the autonomous system. The calculation of the OSPF routing table is based on Dijkstra’s algorithm. 133 Works Done Last Week (11/15/2004-11/21/2004) NBL: Write the switch performance test plan for Accton Review resumes and interview job applicants On-going. 3 applicants were interviewed Write training materials for the bridging and routing testing Research: Find the suitable journals for the submission of the CDN paper World Wide Web Journal Think about the multi-resource scheduling algorithm for Web clusters on-going. 134 Works Planned for This Week (11/22/2004-11/28/2004) NBL: Test DigiChina 7604 Switch Write the switch performance test plan for Accton Final interview to a job applicant Write training materials for the bridging and routing testing Research: Make the slides for the Web QoS patent Present the Web QoS patent at ITRI Think about the multi-resource scheduling algorithm for Web clusters 135 Description (11/22/2004-11/28/2004) A cost is associated with the output side of each OSPF router interface. This cost is configurable by the system administrator. The lower the cost, the more likely the interface is to be used to forward data traffic. When the interface cost (link-state metric) is changed, the OSPF router should recalculate its routing table to construct a new shortest-path tree. 136 Works Done Last Week (11/22/2004-11/28/2004) NBL: Test DigiChina 7604 Switch Write the switch performance test plan for Accton Done. Write training materials for the bridging and routing testing On-going. Final interview to a job applicant IP gateway and RIP conformance tests are done. On-going. Research: Make the slides for the Web QoS patent Present the Web QoS patent at ITRI Done. Done. Think about the multi-resource scheduling algorithm for Web clusters On-going. 137 Works Planned for This Week (11/29/2004-12/05/2004) NBL: Test the conformance of DigiChina and Accton Switches Finish the RFC 2889 Layer 2 Switch Performance Tests Research: Revise the CDN paper Think about the multi-resource scheduling algorithm for Web clusters 138 Description (11/29/2004-12/05/2004) In case of topology change, a router should recalculate routes quickly in the face of topological changes. Here several topology change schemes are discussed, including changing the IP address of an IP interface, disabling a routing process, shutting down an interface, shutting down a router, and removing the network cable from an interface. 139 Works Done Last Week (11/29/2004-12/05/2004) NBL: Test the conformance of DigiChina and Accton Switches Finish the RFC 2889 Layer 2 Switch Performance Tests The OSPF Core of DigiChina and IP Gateway of Accton tests are done. Done Research: Revise the CDN paper On-going. Think about the multi-resource scheduling algorithm for Web clusters On-going 140 Works Planned for This Week (12/06/2004-12/12/2004) NBL: Test the conformance of DigiChina and Accton Switches Perform Layer 3 Forwarding Performance Tests Research: Revise the CDN paper Think about the multi-resource scheduling algorithm for Web clusters 141 Description (12/06/2004-12/12/2004) DVMRP uses a distance vector distributed routing algorithm in order to build per-sourcegroup multicast delivery trees. When an IP multicast datagram arrives, a DVMRP router will forward the datagram to a list of downstream interfaces. As the DVMRP routers at the leaves of the tree begin to receive unwanted multicast traffic, they send DVMRP Prune messages upstream toward the source. 142 Works Done Last Week (12/06/2004-12/12/2004) NBL: Test the conformance of DigiChina and Accton Switches Perform Layer 3 Forwarding Performance Tests The OSPF Extensions of DigiChina and RIP of Accton tests are done. The throughput and latency tests are done. Research: Revise the CDN paper On-going. Think about the multi-resource scheduling algorithm for Web clusters On-gong. 143 Works Planned for This Week (12/13/2004-12/19/2004) NBL: Test the conformance of DigiChina and Accton Switches Perform Layer 3 Forwarding Performance Tests Research: Revise the CDN paper Think about the multi-resource scheduling algorithm for Web clusters 144 Description (12/13/2004-12/19/2004) Once a multicast delivery tree has been pruned back, DVMRP Graft messages are necessary to join new receivers onto the multicast tree. DVMRP Graft messages are sent upstream hopby-hop from the new receiver's first-hop router until a point on the multicast tree is reached. It is necessary to acknowledge each DVMRP Graft message with a DVMRP Graft-Ack message. 145 Works Done Last Week (12/13/2004-12/19/2004) NBL: Test the conformance of DigiChina and Accton Switches Perform Layer 3 Forwarding Performance Tests The IGMP conformance test was done. IP forwarding performance tests were done. Research: Revise the CDN paper Pending. Think about the multi-resource scheduling algorithm for Web clusters On-going. 146 Works Planned for This Week (12/20/2004-12/26/2004) NBL: Test the conformance of DigiChina and Accton Switches Finished Layer 3 Forwarding Performance Tests Perform 48-port performance tests at Accton Research: Find a tutorial-style journal for the submission of the CDN paper 147 Description (12/20/2004-12/26/2004) In case of topology change, a DVMRP router should recalculate the multicast delivery tree quickly in the face of topological changes. Here several topology change schemes are discussed, including changing the IP address of an IP interface, disabling a routing process, shutting down an interface, shutting down a router, and removing the network cable from an interface. 148 Works Done Last Week (12/20/2004-12/26/2004) NBL: Test the conformance of DigiChina and Accton Switches Finished Layer 3 Forwarding Performance Tests Done. Perform 48-port performance tests at Accton The DVMRP conformance test of the DigiChina Switch was done. The tests to Accton ES4649 were done. Not yet the DigiChina Switch. Research: Find a tutorial-style journal for the submission of the CDN paper On-going. 149 Works Planned for This Week (12/27/2004-01/02/2005) NBL: Test the PIM-DM conformance of DigiChina IP QoS performance tests Perform 48-port performance tests at Accton Research: 150 Description (12/27/2004-01/02/2005) DVMRP can forward multicast packets by building the downstream interface list for each packet as it arrives. However, to reduce per packet processing time, the result of the first lookup may be cached in a forwarding table. Then, as routes, downstream dependent neighbors, or group membership change, the cache forwarding table entries must be updated to reflect these changes. 151 Works Done Last Week (12/27/2004-01/02/2005) NBL: Test the PIM-DM conformance of DigiChina IP QoS performance tests Qos Tests havn’t finished yet. But the multicast performance tests were done. Perform 48-port performance tests at Accton Done. Layer 2 performance tests were done. Research: 152 Works Planned for This Week (01/03/2005-01/09/2005) NBL: Test the PIM-DM conformance of Accton ES4649 IP QoS performance testing Announce SIGSWITCH mailing list to the initial members Train the new blood – switch configuration Research: 153 Description (01/03/2005-01/09/2005) PIM-DM assumes that when a source starts sending, all downstream systems want to receive multicast datagrams. Initially, multicast datagrams are flooded to all areas of the network. PIM-DM uses reverse path forwarding to prevent looping of multicast datagrams while flooding. If some areas of the network do not have group members, PIM-DM will prune off the forwarding branch by instantiating prune state. 154 Works Done Last Week (01/03/2005-01/09/2005) NBL: Test the PIM-DM conformance of Accton ES4649 IP QoS performance testing Partially done. Train the new blood – switch configuration Done. Announce SIGSWITCH mailing list to the initial members Done. Done. Plus the layer 2 performance tests were also done. Research: 155 Works Planned for This Week (01/10/2005-01/16/2005) NBL: Test the DHCP client conformance of DigiChina and Accton RIP performance testing Announce SIGSWITCH mailing list to the initial members Train the new blood – IP forwarding performance testing Research: 156 Description (01/10/2005-01/16/2005) Once a multicast delivery tree has been pruned back, PIM-DM Graft messages are necessary to join new receivers onto the multicast tree. PMIDM Graft messages are sent upstream hop-byhop from the new receiver's first-hop router until a point on the multicast tree is reached. It is necessary to acknowledge each DVMRP Graft message with a DVMRP Graft-Ack message. 157 Works Done Last Week (01/10/2005-01/16/2005) NBL: Test the DHCP client conformance of DigiChina and Accton RIP performance testing Done Train the new blood – IP forwarding performance testing Done Announce SIGSWITCH mailing list to the initial members Done Done Research: 158 Works Planned for This Week (01/17/2005-01/23/2005) NBL: Discuss the new case with Attcon OSPF performance testing NBL newsletter Train the new blood – Conformance testing Research: 159 Description (01/17/2005-01/23/2005) PIM-DM is a multicast routing protocol that uses the underlying unicast routing information base to flood multicast datagrams to all multicast routers. If a unicast route changes, a DVMRP router must recalculate the multicast delivery tree. This test verifies that changing the unicast route will affect the PIM-DM delivery tree of multicast traffic. 160 Works Done Last Week (01/17/2005-01/23/2005) NBL: Discuss the new case with Attcon OSPF performance testing Done. The IP multicast conformance tests were announced. Train the new blood – Conformance testing The OSPF link-state database capacity and adjacency capacity tests were done. NBL newsletter Done. Accton will send ES3628C for conformance and performance testing Done. Research: 161 Works Planned for This Week (01/24/2005-01/30/2005) NBL: Assign and guide Accton tests to Jimmy and the part-times OSPF convergence performance testing Attend the NBL security workshop in Taipei Train the new blood – Interoperability testing Train the new blood – IP Multicast performance testing Research: 162 Description (01/24/2005-01/30/2005) In case of topology change, a PIM-DM router should recalculate the multicast delivery tree quickly in the face of topological changes. Here several topology change schemes are discussed, including changing the IP address of an IP interface, disabling a routing process, shutting down an interface, shutting down a router, and removing the network cable from an interface. 163 Works Done Last Week (01/24/2005-01/30/2005) NBL: Assign and guide Accton tests to Jimmy and the part-times OSPF convergence performance testing Only the mixed class throughput was taught. Write the training materials Not yet. Train the new blood – IP Multicast performance testing Done. Train the new blood – Interoperability testing Done. Attend the NBL security workshop in Taipei Done. Done. Research: 164 Works Planned for This Week (01/31/2005-02/13/2005) NBL: Redo Accton ES4649 48-port IP multicast performance testing Train the new blood – Interoperability testing Train the new blood – IP Multicast performance testing Review the NBL documents Research: 165 Description (01/31/2005-02/13/2005) If some areas of the network do not have group members, PIM-DM will prune off the forwarding branch by instantiating prune state. PIM-DM Prune messages are sent towards the upstream neighbor for S to indicate that traffic from S addressed to group G is not desired. In the case of two downstream routers A and B, where A wishes to continue receiving data and B does not, A will send a PIM-DM Join in response to B's PIM-DM Prune to override the Prune. 166 Works Done Last Week (01/31/2005-02/13/2005) NBL: Redo Accton ES4649 48-port IP multicast performance testing Train the new blood – Interoperability testing Done. Review the NBL documents Done. Train the new blood – IP Multicast performance testing Done. Dnoe. Research: 167 Works Planned for This Week (02/14/2005-02/20/2005) NBL: Write test procedures for performance testing Review the performance test results of Accton ES3628C Research: Qualifying exam 168 Description (02/14/2005-02/20/2005) Once a multicast delivery tree has been pruned back, PIM-DM Graft messages are necessary to join new receivers onto the multicast tree. PMIDM Graft messages are sent upstream hop-byhop from the new receiver's first-hop router until a point on the multicast tree is reached. It is necessary to acknowledge each DVMRP Graft message with a DVMRP Graft-Ack message. 169 Works Done Last Week (02/14/2005-02/20/2005) NBL: Write test procedures for performance testing Review the performance test results of Accton ES3628C 78% done. The review for the layer 2 and layer 3 performance tests were done. Research: Qualifying exam Done. 170 Works Planned for This Week (02/21/2005-02/27/2005) NBL: Write test procedures for performance testing Write a proposal for the cooperation with Accton Review the multicast performance test results of Accton ES3628C Research: Review and revise the Web QoS patent. Discuss QoS for Web services with ITRI on Thursday 171 Description (02/21/2005-02/27/2005) If some areas of the network do not have group members, PIM-DM will prune off the forwarding branch by instantiating prune state. PIM-DM Prune messages are sent towards the upstream neighbor for S to indicate that traffic from S addressed to group G is not desired. In the case of two downstream routers A and B, where A wishes to continue receiving data and B does not, A will send a PIM-DM Join in response to B's PIM-DM Prune to override the Prune. 172