Google_Yaoshen

advertisement
Tufts University
Virtual Machine Usage
in Cloud Computing in Google
EE-126
Yaoshen Yuan
www.company.com
Tufts University
Google Cloud Computing Platform
• SaaS (Software as a Service)
– Clients can download software and other resources or
create documents and save resources through SaaS.
• PaaS (Platform as a Service)
– provides clients with the platform that allow them to
deploy the virtual development environment.
• IaaS (Infrastructure as a Service)
– shares the Internet infrastructure and consumers have
the ability to configure the operating system, storage,
applications.
www.company.com
Tufts University
Diagram for Platform
IaaS
PaaS
SaaS
Application
Application
Application
Data
Data
Data
Runtime
Runtime
Runtime
Middleware
Middleware
Middleware
O/S
O/S
O/S
Virtualization
Virtualization
Virtualization
Server
Server
Server
Storage
Storage
Storage
Networking
Networking
Networking
www.company.com
Tufts University
Analysis of VM usage in Google Cloud
• Transmission Latency
• Traffic Analysis
• Reliability
www.company.com
Tufts University
Transmission Latency
•
1
2
3
4
5
6
7
8
9
10
11
12
13
24.05
22.31
1.34
60.57
50.45
53.31
33.20
35.91
33.75
36.31
41.26
45.59
-33.37
120.5
114.2
103.7
27.19
3.28
-6.26
-80.01
-81.54
-84.75
-95.32
-95.87
-121.1
-70.73
www.company.com
Tufts University
Transmission Latency
virtual machine will access the nearest
data center, so the distance should be
www.company.com
Tufts University
Transmission Latency
P reflects the advantage of cloud computing
compared to the server station model
www.company.com
Tufts University
Traffic Analysis
the augmentation of the number of request a
data center receive increases the network latency,
0.1
0.1
so it0.1is necessary
to consider
the network traffic
LONGITUDE
LATITUDE
data center1
0.1
data center3
0.1
0.2
0.1
data center2
0.1
0.1
www.company.com
Tufts University
Traffic Analysis
center
1
2
3
4
5
6
7
8
9
10
11
12
13
mean
0.0778
0.0206
0.1547
0.1411
0.0552
0.0446
0.0269
0.0098
0.0065
0.0169
0.0144
0.1136
0.3178
mean traffic weight of each data center under the condition that
request is produced randomly over the world during a day
www.company.com
Tufts University
Reliability
it is important that when one or some of the data center
collapse, VM instances can still access their resources
www.company.com
Tufts University
Reliability
www.company.com
Tufts University
Reliability
www.company.com
Tufts University
Reliability
www.company.com
Tufts University
Reliability
k
0
1
2
3
4
5
6
7
8
9
10
11
12
43965
67.93
57513
98.49
58482
13.53
58659
02.77
59176
84.76
59360
62.29
59516
09.23
66690
10.87
67462
55.67
70307
87.34
89005
97.50
98420
86.84
10008
160.72
www.company.com
Tufts University
Reliability
www.company.com
Tufts University
Conclusion
 Because of the lack of data of real channel
connecting the world, the model (using straight line
in sphere to replace channel) used to analyze is not
accurate.
 Model built under the condition that resources of
one VM instance are saved in all data center.
 Less latency, higher traffic tolerance, higher
reliability.
 Building server on Google Cloud using VM instance
is sensible when large Page View (PV) is estimated.
www.company.com
Tufts University
REFERENCE
[1] Niyato D. Optimization-based virtual machine manager for private cloud
computing[C]//Cloud Computing Technology and Science (CloudCom),
2011 IEEE Third International Conference on. IEEE, 2011: 99-106.
[2] Rajan S, Jairath A. Cloud computing: The fifth generation of
computing[C]//Communication Systems and Network Technologies
(CSNT), 2011 International Conference on. IEEE, 2011: 665-667.
[3] Ye K, Huang D, Jiang X, et al. Virtual machine based energy-efficient
data center architecture for cloud computing: a performance
perspective[C]//Proceedings of the 2010 IEEE/ACM Int'l Conference on
Green Computing and Communications & Int'l Conference on Cyber,
Physical and Social Computing. IEEE Computer Society, 2010: 171-178.
[4] Savu L. Cloud computing: Deployment models, delivery models, risks
and research challenges[J]. Computer, 2011.
[5] Managed VM, https://cloud.google.com/appengine/docs/managed-vms/
[6] Shang Z, Chen W, Ma Q, et al. Design and implementation of server
cluster dynamic load balancing based on OpenFlow[C]//Awareness
Science and Technology and Ubi-Media Computing (iCAST-UMEDIA),
2013 International Joint Conference on. IEEE, 2013: 691-697.
[7] Google Data Center
http://www.google.com/about/datacenters/inside/locations/index.html
www.company.com
Tufts University
www.company.com
Download