Evaluating the Cost-Benefit of Using Cloud Computing to Extend the

advertisement
Evaluating the Cost-Benefit of
Using Cloud Computing to Extend
the Capacity of Clusters
Presenter: Xiaoyu Sun
Cluster Computing
Users have to know cluster very well
administrative privileges
What is Cloud Computing?
Cloud computing provides computation,
software, data access, and storage resources
without requiring cloud users to know the
location and other details of the computing
infrastructure.
Characteristics of Cloud Computing

Empowerment

Agility





Users control resource by themselves not by a centralized IT service
users' ability to re-provision technological infrastructure resources.
Application Programing Interface
Cost
Device and Location Independence

enable users to access systems using a web browser regardless of their location or what device they are using

Virtualization


Reliability and Scalability
Performance

Security

Maintenance




servers and storage devices to be shared and utilization be increased
Monitor by web services as the system interface
providers are able to devote resources to solving security issues that many customers cannot afford
Applications don’t need to be installed on each user's computer and can be accessed from different places
Purpose
 Describe a system that enables an organization to
augment its computing infrastructure by allocating
resources from a Cloud provider.
 Provide various scheduling strategies that aim to minimize
the cost of utilizing resources from the Cloud provider.
 Evaluate the proposed strategies, considering different
performance metrics; namely average weighted response
time, job slowdown, number of deadline violations,
number of jobs rejected, and the
using the Cloud.
money spent for
Cloud Computing
Strategy sets
Figure 1:The resource provisioning scenario
Backfilling Policies
 Conservative
 each request is scheduled when it arrives in the system, and requests are allowed
to jump ahead in the queue if they do not delay the execution of other requests.
 Aggressive
 Only the request at the head of the waiting queue called the pivot is granted a
reservation. Other requests are allowed to move ahead in the queue if they do not
delay the pivot.
 Selective
 Requests are given reservations if they have waited long enough in the queue.
Long enough is determined by the requests’ expansion factor:
requests.
Xfactor = (wait time + run time)/run time (1)
The threshold is given by the average slowdown of previously completed
Strategy Sets
 Naïve:
 Both site and cloud schedulers use Conservative
backfilling to schedule the requests
 The redirection algorithm is executed at the arrival of
each job at the site
 Use cloud provider when the request cannot start
immediately on local cluster
Strategy Sets
 Shortest Queue:
 Aggressive backfilling
 First-Come-First-Served (FCFS) manner
 At the arrival or complete of each job at the site
 Compute the ratio of number of VMs required by
requests to the number of VMS available
 Redirect request if cloud provider’s number is smaller
Strategy Sets
 Weighted Queue:
 Aggressive backfilling
 First-Come-First-Served (FCFS) manner
 Number of VMs that can be borrowed from cloud
provider is the number of VMs required by requests
minus VMs in use
Strategy Sets
 Selective
 Selective backfilling
 Compute the ratio of number of VMs required by
requests to the number of VMS available
 When the request’s xFactor exceeds the threshold, the
scheduler makes a reservation at the place that
provides the earliest start time.
Experiments
 Simulation of two-month-long periods
 SDSC Blue Horizon machine with 144 nodes
 Number of VMs
 Price of a virtual machine per hour
 Amazon EC2’s small instance: US $0.10
 Network and storage are not considered
 Values are average of 5 simulation runs
Performance Metrics
 Average Weighted Response Time(AWRT) of site k:
AWRTk =
å
jÎTk
p j × m j × (ct j - st j )
å p ×m
j
jÎTk
(2)
j
 Tk: requests submitted to site k
 Pj: the runtime of request j
 mj: the number of VMs required by request j
 ctj: request j’s completion time
 stj: the submission time of request j
Performance Metrics
 Performance Improvement Cost of a strategy set st:
Amount _ spend
PICst =
* AWRTst
AWRTbase - AWRTst
(3)
 Amount spent is the amount spent running virtual machines on the Cloud provider
 AWRTbase is the AWRT achieved by a base strategy(FCFS with aggressive
backfilling) that schedules requests using only the site's resources
 AWRTst is the AWRT reached by the strategy st when Cloud resources are also
utilized.
Performance Improvement Cost
 Using Lublin99's model to generate different workloads:
 Umed: the mean number of virtual machines required by a
request to log2m-umed where m is the maximum number of
virtual machines allowed in the system, from 1.5 to 3.5.
 Barr: the inter-arrival time of requests at rush hours, from 0.45
to 0.55 .
 PB: the proportion p of the first gamma in Lublin99's model is given
by p = pa * nodes + PB, from 0.5 to 1.0.
Performance Improvement Cost
These three graphs show the site's utilization using the base aggressive
backfilling strategy without Cloud resources
The larger the value of Umed, the smaller the requests.
The larger the value of PB, the smaller the duration
of the requests
Performance Improvement Cost
Requests’ size
Requests’ arrive time
Requests’ duration
Deadline Constrained Applications
 Users may have stringent requirement on when the
virtual machines are required
 Deadline constrained requests have:
 Ready time
 Duration
 Deadline
 Cost of using Cloud resources used to meet requests’
deadlines and decrease the number of deadline
violations and request rejections
Deadline Aware Strategies
 Conservative
 both local site and Cloud schedule requests using conservative
backfilling.
 Places a request where it achieves the best start time
 If rejections are allowed and deadline cannot be met, reject the request
 Aggressive




both local site and Cloud use aggressive backfilling to schedule requests
Earliest Deadline First
If request deadlines are broken in the local cluster, try the cloud provider
If rejections are allowed and deadlines are broken, reject the request
Cost of Reducing Deadline Violations
 The non-violation cost is given by:
Amount _ spentst
non - violation costst =
(4)
violbase - violst
 Where:
 Amount_spentst: amount spent with Cloud resources
 Violbase: the number of deadline violations under the base
strategy set (aggressive backfilling and an Earliest Deadline
First manner)
 Violst:the number of deadline violations under the evaluated
strategy set
Deadline calculation
 The deadline calculation is given by:
 Where:
 stj: the request j's submission time
 ctj: the completion time.
 taj: the difference between the request's completion and submission times.
 sf : a stringency factor that indicates how urgent the deadlines are.
Cost of Reducing Deadline Violations
sf=0.9
sf=1.3
sf=1.7
Cost of Reducing Deadline Violations
Tight deadlines
Normal deadlines
Relaxed deadlines
Cost to Reduce Job Rejections:
Aggressive Strategy Set
Cost to Reduce Job Rejections:
Aggressive Strategy Set
Conclusions
 Different strategy sets can yield different ratios of
performance improvement to money spent
 Naïve strategy has a higher performance improvement
cost
 Selective strategy provides a good ratio of money spent
to job slowdown improvement
 Using cloud provider to meet job deadlines
 Less than $3,000 were spent to keep the number of
rejections close to zero
Download
Related flashcards

Theory of computation

16 cards

Computer science

25 cards

MSN

28 cards

Software

43 cards

System software

24 cards

Create Flashcards