Cost-Benefit Analysis of Cloud Computing versus Desktop Grids - - PowerPoint PPT Presentation

▶
cost benefit analysis of cloud computing versus desktop
SMART_READER_LITE
LIVE PREVIEW

Cost-Benefit Analysis of Cloud Computing versus Desktop Grids - - PowerPoint PPT Presentation

Apr 20, 2023 460 likes •1.16k views

Cost-Benefit Analysis of Cloud Computing versus Desktop Grids Derrick Kondo, Bahman Javadi, Paul Malcot, Franck Cappello INRIA, France David P. Anderson UC Berkeley, USA Cloud Background Vision Hide complexity of hardware and

slide-1
SLIDE 1

Cost-Benefit Analysis of Cloud Computing versus Desktop Grids

Derrick Kondo, Bahman Javadi, Paul Malécot, Franck Cappello INRIA, France David P. Anderson UC Berkeley, USA

slide-2
SLIDE 2

Cloud Background

  • Vision
  • Hide complexity of hardware and

software management from a user by

  • ffering computing as a service
  • Benefits
  • Pay as you go
  • Scale up or down dynamically
  • No hardware management, less software

management

slide-3
SLIDE 3

Platform Performance vs. Costs

Cost Performance

low low high high

slide-4
SLIDE 4

Platform Performance vs. Costs

Cost Performance

Cluster Grid

Supercomputer

low low high high

slide-5
SLIDE 5

Platform Performance vs. Costs

Cost Performance

Cluster Grid

Supercomputer

low low high high

Desktop Grids (DG)

slide-6
SLIDE 6

Platform Performance vs. Costs

Cost Performance

Cluster Grid

Supercomputer

low low high high

Desktop Grids (DG)

Cloud

What is the relationship?

slide-7
SLIDE 7

Platform Performance vs. Costs

Cost Performance

Cluster Grid

Supercomputer

low low high high

Desktop Grids (DG)

Cloud

What is the relationship?

DG + Cloud?

slide-8
SLIDE 8

Outline

  • Performance tradeoffs
  • Monetary tradeoffs
  • Client hosting
  • Server hosting
slide-9
SLIDE 9

Apples to Apples

Loosely-coupled, high-throughput, compute-intensive applications Tightly-coupled, data-intensive real-time applications

low complexity high complexity

slide-10
SLIDE 10

Apples to Apples

Loosely-coupled, high-throughput, compute-intensive applications Tightly-coupled, data-intensive real-time applications

low complexity high complexity

DG’s

slide-11
SLIDE 11

Apples to Apples

Loosely-coupled, high-throughput, compute-intensive applications Tightly-coupled, data-intensive real-time applications

low complexity high complexity

DG’s Clouds

slide-12
SLIDE 12

Apples to Apples

Loosely-coupled, high-throughput, compute-intensive applications Tightly-coupled, data-intensive real-time applications

low complexity high complexity

DG’s Clouds

Comparison assuming embarrassingly parallel, compute-intensive applications

slide-13
SLIDE 13

Method

  • Use real performance measurements
  • Exported BOINC (middleware for desktop grids)

project data

  • Use real costs
  • Large/small BOINC projects (SETI@home /

XtremLab)

  • Amazon Elastic Computing Cloud (EC2)
slide-14
SLIDE 14

Stages of Project & Application

Platform Construction

slide-15
SLIDE 15

Stages of Project & Application

Platform Construction Application Deployment

slide-16
SLIDE 16

Stages of Project & Application

Platform Construction Application Deployment Application Execution

slide-17
SLIDE 17

Stages of Project & Application

Platform Construction Application Deployment Application Execution Application Completion

slide-18
SLIDE 18

Platform Construction Application Deployment Application Execution Application Completion

How long before I get X TeraFLOPS?

slide-19
SLIDE 19

Platform Construction Application Deployment Application Execution Application Completion

How long before I get X TeraFLOPS?

5 10 15 20 25 30 10

3

10

4

10

5

10

6

Months for registration Number of cloud nodes 5 10 15 20 25 30 10

1

10

2

10

3

TeraFLOPS

slide-20
SLIDE 20

Platform Construction Application Deployment Application Execution Application Completion

How long before I get X TeraFLOPS?

5 10 15 20 25 30 10

3

10

4

10

5

10

6

Months for registration Number of cloud nodes 5 10 15 20 25 30 10

1

10

2

10

3

TeraFLOPS

Can get over 20 TeraFLOPS within 6 months

slide-21
SLIDE 21

Platform Construction Application Deployment Application Execution Application Completion

How long before I get X TeraFLOPS?

5 10 15 20 25 30 10

3

10

4

10

5

10

6

Months for registration Number of cloud nodes 5 10 15 20 25 30 10

1

10

2

10

3

TeraFLOPS

Can get over 20 TeraFLOPS within 6 months

Strategy: Add to BOINC project list Press releases Forum Announcements Google Ad Sense Respond to users (leverage volunteers)

slide-22
SLIDE 22

Platform Construction Application Deployment Application Execution Application Completion

How long to deploy my batch of tasks needing faster response time?

slide-23
SLIDE 23

Platform Construction Application Deployment Application Execution Application Completion

How long to deploy my batch of tasks needing faster response time?

!"

#

!"

$

!"

%

!"

!!

!"

"

!"

!

!"

&

!"

'

()*+,-./0.1/2)34,,-.3/5,6 7,82/9*,34.4:*,.;*:3)4,6< . . !"".4=6>6 !""".4=6>6 !"""".4=6>6

slide-24
SLIDE 24

Platform Construction Application Deployment Application Execution Application Completion

How long to deploy my batch of tasks needing faster response time?

!"

#

!"

$

!"

%

!"

!!

!"

"

!"

!

!"

&

!"

'

()*+,-./0.1/2)34,,-.3/5,6 7,82/9*,34.4:*,.;*:3)4,6< . . !"".4=6>6 !""".4=6>6 !"""".4=6>6

For 1000 tasks, ~10 minutes with 105 hosts

slide-25
SLIDE 25

Platform Construction Application Deployment Application Execution Application Completion

How long to deploy my batch of tasks needing faster response time?

!"

#

!"

$

!"

%

!"

!!

!"

"

!"

!

!"

&

!"

'

()*+,-./0.1/2)34,,-.3/5,6 7,82/9*,34.4:*,.;*:3)4,6< . . !"".4=6>6 !""".4=6>6 !"""".4=6>6

For 1000 tasks, ~10 minutes with 105 hosts

Strategy: Specify lower latency bounds [Heien et al.]

slide-26
SLIDE 26

Platform Construction Application Deployment Application Execution Application Completion

slide-27
SLIDE 27

Platform Construction Application Deployment Application Execution Application Completion

How many volunteer nodes are equivalent to 1 cloud node?

slide-28
SLIDE 28

Platform Construction Application Deployment Application Execution Application Completion

How many volunteer nodes are equivalent to 1 cloud node?

slide-29
SLIDE 29

Platform Construction Application Deployment Application Execution Application Completion

How many volunteer nodes are equivalent to 1 cloud node?

2.8 active volunteer hosts per 1 cloud node. (Total performance still

  • rders of magnitude better)
slide-30
SLIDE 30

Platform Construction Application Deployment Application Execution Application Completion

How many volunteer nodes are equivalent to 1 cloud node?

2.8 active volunteer hosts per 1 cloud node. (Total performance still

  • rders of magnitude better)

Strategy: Use statistical prediction of availability

slide-31
SLIDE 31

Platform Construction Application Deployment Application Execution Application Completion

slide-32
SLIDE 32

Platform Construction Application Deployment Application Execution Application Completion

How long should I wait for task completion?

slide-33
SLIDE 33

Platform Construction Application Deployment Application Execution Application Completion

How long should I wait for task completion?

slide-34
SLIDE 34

Platform Construction Application Deployment Application Execution Application Completion

How long should I wait for task completion?

Median project latency bound: 9 days for 3.7 hour work unit (on 3GHz host). Ratio of lat. bound / exec time > 5. Good success rates: 96.1%

  • f WCG tasks met out of 227,000 tasks
slide-35
SLIDE 35

Platform Construction Application Deployment Application Execution Application Completion

How long should I wait for task completion?

Strategy: See BOINC Catalog for typical deadlines and compute/comm/mem ratios.

Median project latency bound: 9 days for 3.7 hour work unit (on 3GHz host). Ratio of lat. bound / exec time > 5. Good success rates: 96.1%

  • f WCG tasks met out of 227,000 tasks
slide-36
SLIDE 36

Monetary Tradeoffs

  • Client hosting on cloud
  • Not worth it and never will
  • Server hosting on the cloud
  • Possible solution
slide-37
SLIDE 37

Monthly Project Costs

slide-38
SLIDE 38

EC2 Pricing

slide-39
SLIDE 39

What about Client Hosting on the Cloud?

slide-40
SLIDE 40

Cost of Clouds versus DG

How long until DG’s more cost effective than Clouds?

slide-41
SLIDE 41

Cost of Clouds versus DG

! " #! #" $! $" %! #!

!#&

#!

!#"

#!

!#'

#!

!#%

#!

!#$

#!

!##

()*+,-./01,/230456 7,8./9,)/:;<( / / =.),>;0? @ABCDE*>, A7$ F#'!'/G*HI8.,,)/8*3,5 2'J&/-H*I3/8*3,56 '"&$/G*HI8.,,)/8*3,5 2#&#$/-H*I3/8*3,56

How long until DG’s more cost effective than Clouds?

slide-42
SLIDE 42

Cost of Clouds versus DG

! " #! #" $! $" %! #!

!#&

#!

!#"

#!

!#'

#!

!#%

#!

!#$

#!

!##

()*+,-./01,/230456 7,8./9,)/:;<( / / =.),>;0? @ABCDE*>, A7$ F#'!'/G*HI8.,,)/8*3,5 2'J&/-H*I3/8*3,56 '"&$/G*HI8.,,)/8*3,5 2#&#$/-H*I3/8*3,56

How long until DG’s more cost effective than Clouds?

After ~13 days

slide-43
SLIDE 43

What are total costs

  • ver time?

Cost of Clouds versus DG (2)

slide-44
SLIDE 44

5 10 15 20 1 2 3 4 5 6 7 8 x 10

4

Platform age (days) Total cost EC2 100n EC2 1000n EC2 10000n XtremLab SETI@home

What are total costs

  • ver time?

Cost of Clouds versus DG (2)

slide-45
SLIDE 45

5 10 15 20 1 2 3 4 5 6 7 8 x 10

4

Platform age (days) Total cost EC2 100n EC2 1000n EC2 10000n XtremLab SETI@home

What are total costs

  • ver time?

Cost of Clouds versus DG (2)

After 3 days, DG is cheaper

slide-46
SLIDE 46

Cost of Clouds versus DG (3)

How many months

  • f DG can

X months of Cloud buy me?

slide-47
SLIDE 47

5 10 15 20 25 500 1000 1500 2000 2500 3000 3500 Months with EC2 Months with VC Xtremlab / 1000n EC2 SETI@home / 1000n EC2 Xtremlab / 10000n EC2 SETI@home / 10000n EC2

Cost of Clouds versus DG (3)

How many months

  • f DG can

X months of Cloud buy me?

slide-48
SLIDE 48

5 10 15 20 25 500 1000 1500 2000 2500 3000 3500 Months with EC2 Months with VC Xtremlab / 1000n EC2 SETI@home / 1000n EC2 Xtremlab / 10000n EC2 SETI@home / 10000n EC2

Cost of Clouds versus DG (3)

How many months

  • f DG can

X months of Cloud buy me?

12 months of Cloud can pay for > 125 DG years. DG way more sustainable.

slide-49
SLIDE 49

Equivalent Clouds given BOINC Costs

Given BOINC costs, what size Cloud can I buy?

slide-50
SLIDE 50

Equivalent Clouds given BOINC Costs

Given BOINC costs, what size Cloud can I buy?

slide-51
SLIDE 51

Equivalent Clouds given BOINC Costs

Given BOINC costs, what size Cloud can I buy?

1.75 TeraFLOPS, 7 TeraBytes .75 TeraFLOPS, 3 TeraBytes

slide-52
SLIDE 52

Equivalent Clouds given BOINC Costs

Given BOINC costs, what size Cloud can I buy?

2 orders

  • f magnitude

lower than BOINC

1.75 TeraFLOPS, 7 TeraBytes .75 TeraFLOPS, 3 TeraBytes

slide-53
SLIDE 53

What about Server Hosting on the Cloud?

slide-54
SLIDE 54

Number of Hosts over Time

slide-55
SLIDE 55

Number of Hosts over Time

Load variation exits w/ publicity, projects run out

  • f work, etc.

Clouds take care

  • f server

management. Potential to exploit clouds.

slide-56
SLIDE 56

Server Costs on a Cloud

How much to host BOINC server on cloud?

slide-57
SLIDE 57

Server Costs on a Cloud

How much to host BOINC server on cloud?

slide-58
SLIDE 58

Server Costs on a Cloud

How much to host BOINC server on cloud?

(versus 43K +12K/month) (versus {4,1}K + {5,1}K/month)

slide-59
SLIDE 59

Server Costs on a Cloud

1.7 - 16.5 times cheaper to host on Cloud, but bandwidth is expensive. How much to host BOINC server on cloud?

(versus 43K +12K/month) (versus {4,1}K + {5,1}K/month)

slide-60
SLIDE 60

Cloud Resources with Given Budget

How big of a server can I support with given budget?

slide-61
SLIDE 61

Cloud Resources with Given Budget

How big of a server can I support with given budget?

slide-62
SLIDE 62

Cloud Resources with Given Budget

How big of a server can I support with given budget? 4 Mbps, 2 TB 12 Mbps, 2 TB

slide-63
SLIDE 63

Cloud Resources with Given Budget

Many project servers are sustainable on the cloud How big of a server can I support with given budget? 4 Mbps, 2 TB 12 Mbps, 2 TB

slide-64
SLIDE 64

Summary

slide-65
SLIDE 65

Summary

  • Performance tradeoffs
  • 20 DG TeraFLOPS within 6 months
slide-66
SLIDE 66

Summary

  • Performance tradeoffs
  • 20 DG TeraFLOPS within 6 months
  • Monetary tradeoffs
  • Client hosting
  • After 13 days, DG more cost effective
  • Server hosting
  • DG server on cloud is cost-effective
  • Best for small-medium sized projects
  • Savings of at least 40%
  • http://mescal.imag.fr/membres/derrick.kondo/cloud_calc.xlsx
slide-67
SLIDE 67

Summary

  • Performance tradeoffs
  • 20 DG TeraFLOPS within 6 months
  • Monetary tradeoffs
  • Client hosting
  • After 13 days, DG more cost effective
  • Server hosting
  • DG server on cloud is cost-effective
  • Best for small-medium sized projects
  • Savings of at least 40%
  • http://mescal.imag.fr/membres/derrick.kondo/cloud_calc.xlsx
  • BOINC server image for EC2 / Xen available
  • http://boinc.berkeley.edu/trac/wiki/CloudServer
slide-68
SLIDE 68

Future Work

  • Clouds@home
  • Statistical and predictive methods for

ensuring host availability

  • Lightweight mechanisms for virtual

machine migration

slide-69
SLIDE 69
  • wu flop calc
  • bw in cloud
  • how is cloud equiv calculated