Towards Deploying Decommissioned Mobile Devices as Cheap ... · Towards Deploying Decommissioned...

20
Towards Deploying Decommissioned Mobile Devices as Cheap Energy-Efficient Compute Nodes Mohammad Shahrad and David Wentzlaff Monday, July 10, 2017

Transcript of Towards Deploying Decommissioned Mobile Devices as Cheap ... · Towards Deploying Decommissioned...

Page 1: Towards Deploying Decommissioned Mobile Devices as Cheap ... · Towards Deploying Decommissioned Mobile Devices as Cheap Energy-Efficient Compute Nodes Mohammad Shahrad and David

Towards Deploying Decommissioned Mobile Devices as Cheap Energy-Efficient Compute Nodes

Mohammad Shahrad and David Wentzlaff

Monday, July 10, 2017

Page 2: Towards Deploying Decommissioned Mobile Devices as Cheap ... · Towards Deploying Decommissioned Mobile Devices as Cheap Energy-Efficient Compute Nodes Mohammad Shahrad and David

Growing Dominance of Smartphones

2

[Gartner, IDC, Zuberbühler Associates AG][Ericsson Mobility Report 2016]

2015 2021

3.2 billion

6.3 billion

Smartphones

PCsTablets

Page 3: Towards Deploying Decommissioned Mobile Devices as Cheap ... · Towards Deploying Decommissioned Mobile Devices as Cheap Energy-Efficient Compute Nodes Mohammad Shahrad and David

Decommissioned Smartphones

3

Physical Damage Expanding OSs

New Features Fashion

Bulk Recycling

Electronic Waste

Refurbish

Page 4: Towards Deploying Decommissioned Mobile Devices as Cheap ... · Towards Deploying Decommissioned Mobile Devices as Cheap Energy-Efficient Compute Nodes Mohammad Shahrad and David

Reusing smartphones as compute nodes?

4

Using the computational capacity of decommissioned devices

Page 5: Towards Deploying Decommissioned Mobile Devices as Cheap ... · Towards Deploying Decommissioned Mobile Devices as Cheap Energy-Efficient Compute Nodes Mohammad Shahrad and David

Outline

5

Why using mobile SoCs now?

How to integrate mobile devices into data center fabric?

What to do with such mobile devices in data centers?

Does such deployment make economic sense?

Page 6: Towards Deploying Decommissioned Mobile Devices as Cheap ... · Towards Deploying Decommissioned Mobile Devices as Cheap Energy-Efficient Compute Nodes Mohammad Shahrad and David

Trend 1: Commodity vs. Mobile Performance

6

[Rajovic et al., Supercomputing with commodity CPUs: Are mobile SoCs ready for HPC?, SC ’13]

Page 7: Towards Deploying Decommissioned Mobile Devices as Cheap ... · Towards Deploying Decommissioned Mobile Devices as Cheap Energy-Efficient Compute Nodes Mohammad Shahrad and David

Trend 2: Diminishing Performance Growth

7

End of Moore’s Law is starting to kick in for mobile SoCs.

The performance gap between a new vs. a 3-year old phone is decreasing.

8,000

80,000

CPU

Mar

k R

atin

g

Release Date

Page 8: Towards Deploying Decommissioned Mobile Devices as Cheap ... · Towards Deploying Decommissioned Mobile Devices as Cheap Energy-Efficient Compute Nodes Mohammad Shahrad and David

Extending Effective Lifetime to Improve Carbon Footprint

8

[Ercan et al., Life cycle assessment of a smartphone, ICT4S ’16]

Page 9: Towards Deploying Decommissioned Mobile Devices as Cheap ... · Towards Deploying Decommissioned Mobile Devices as Cheap Energy-Efficient Compute Nodes Mohammad Shahrad and David

Proposed Integration Fabric

9

Mobile Device Cage(180 × 80 × 9 mm)

FansRouter

Power Supply

2U Height

84 × 470~~ SoC TDP = 3.5 W

Page 10: Towards Deploying Decommissioned Mobile Devices as Cheap ... · Towards Deploying Decommissioned Mobile Devices as Cheap Energy-Efficient Compute Nodes Mohammad Shahrad and David

Networking

10

1. USB Tree + Master Node

101010

(a) USB Tree A (b) USB Tree B

1010

Master Node Master Node

Smartphones

3. Intra-server WiFi

Supply

Low SNR and high congestion.

Low BW High

Energy

High Latency

2. USB On-The-Go (OTG)

USB OTG Ethernet Ethernet Switch

Upstream Network

Ethernet Switch Cost (price+space)

High BW per DeviceNo Need for a Master Node

Establish virtual Ethernet through master nodeMaster becomes single point of failure

Page 11: Towards Deploying Decommissioned Mobile Devices as Cheap ... · Towards Deploying Decommissioned Mobile Devices as Cheap Energy-Efficient Compute Nodes Mohammad Shahrad and David

Mobile Batteries as Distributed UPSs

Batteries could be used to perform power capping in data centers.

Mobile batteries are designed forhigh-density energy storage.

Average mobile battery: > 3100mAh After 3 years (15% dep/yr): >1900mAh

84 device per server: ~160Ah (4x-8x the typical UPS density)

11

Time

Pow

er D

raw

Power Cap

Battery Power Draw

Battery Recharge

Page 12: Towards Deploying Decommissioned Mobile Devices as Cheap ... · Towards Deploying Decommissioned Mobile Devices as Cheap Energy-Efficient Compute Nodes Mohammad Shahrad and David

Targeted Applications?

12

Non-CPU Elements of Mobile SoC Lower Reliability

Lower CPU PerformanceHigh Relative I/O Bandwidth

Page 13: Towards Deploying Decommissioned Mobile Devices as Cheap ... · Towards Deploying Decommissioned Mobile Devices as Cheap Energy-Efficient Compute Nodes Mohammad Shahrad and David

I/O-intensive Application

13

Smaller cores generally have better I/O to compute ratio.

Mobile cores have better performance-per-Watt.

Prior examples:

Fast Array of Wimpy Nodes [Andersen et al., SOSP ’09] • Using low-end embedded CPUs in a key-value store system. • Two orders of magnitude more queries per Joule.

Web Search Using Mobile Cores [Reddi et al., ISCA ’10] • Comparing Atom to Xeon processors or perform web search. • Better energy efficiency • Worse QoS

Page 14: Towards Deploying Decommissioned Mobile Devices as Cheap ... · Towards Deploying Decommissioned Mobile Devices as Cheap Energy-Efficient Compute Nodes Mohammad Shahrad and David

Low-end VM Provisioning

14

t2.nano

t2.micro

f1-micro

g1-small Multiple tiny burstable instances can be provisioned on each mobile device.

Each mobile device (on average): • > 2GB of memory • > 5 cores

vCPU Mem (GB)11

0.51

vCPU Mem (GB)0.20.5

0.61.7

Page 15: Towards Deploying Decommissioned Mobile Devices as Cheap ... · Towards Deploying Decommissioned Mobile Devices as Cheap Energy-Efficient Compute Nodes Mohammad Shahrad and David

GPU-accelerated Dwarfs

15

Lower-reliability Nodes

•Current GPU-accelerated IaaS instances use beefy GPUs (e.g. EC2 P2 instances)

•Mobile SoCs are equipped with smaller GPUs

•Enabling low-end GPU acceleration for small VMs

•Dynamic reliability platforms use resource heterogeneity for cost saving while meeting SLAs

•Mobile devices can be used as low-cost low-reliability nodes

Page 16: Towards Deploying Decommissioned Mobile Devices as Cheap ... · Towards Deploying Decommissioned Mobile Devices as Cheap Energy-Efficient Compute Nodes Mohammad Shahrad and David

Does it make economic sense?

16

Page 17: Towards Deploying Decommissioned Mobile Devices as Cheap ... · Towards Deploying Decommissioned Mobile Devices as Cheap Energy-Efficient Compute Nodes Mohammad Shahrad and David

TCO Comparison with a Similar-density Server

17

Two simplifying assumptions:

1.Parallelizable workload (using Geekbench 4 cross-platform benchmark)

2.Considered eBay retail price for used phones (could be much cheaper)

TCO: Total Cost of Ownership

Page 18: Towards Deploying Decommissioned Mobile Devices as Cheap ... · Towards Deploying Decommissioned Mobile Devices as Cheap Energy-Efficient Compute Nodes Mohammad Shahrad and David

6 12 18 24 30 36 42 48 54 60

Lifetime (month)

400

600

800

1000

1200

1400

Month

ly T

CO

($)

A, A=45%

B, B

=45%

6 12 18 24 30 36 42 48 54 60

Lifetime (month)

400

600

800

1000

1200

1400

Month

ly T

CO

($)

A, A=45%

B, B

=77.5%

6 12 18 24 30 36 42 48 54 60

Lifetime (month)

0

1000

2000

3000

4000

5000

6000

Month

ly T

CO

($)

A, A=100%

B, B

=100%

A=45%,

B=45%

TCOB

<TCOA

6 12 18 24 30 36 42 48 54 60

Server A Lifetime (month)

6

12

18

24

30

36

42

48

54

60

Serv

er

B L

ifetim

e (

month

)

A=45%,

B=77.5%

TCO B<TCO A

TCO A<TCO B

6 12 18 24 30 36 42 48 54 60

Server A Lifetime (month)

6

12

18

24

30

36

42

48

54

60

Serv

er

B L

ifetim

e (

month

)

A=100%,

B=100%

TCO B<TCO A

TCO A<TCO B

6 12 18 24 30 36 42 48 54 60

Server A Lifetime (month)

6

12

18

24

30

36

42

48

54

60

Serv

er

B L

ifetim

e (

month

)6 12 18 24 30 36 42 48 54 60

Lifetime (month)

400

600

800

1000

1200

1400

Month

ly T

CO

($)

A, A=45%

B, B

=45%

6 12 18 24 30 36 42 48 54 60

Lifetime (month)

400

600

800

1000

1200

1400

Month

ly T

CO

($)

A, A=45%

B, B

=77.5%

6 12 18 24 30 36 42 48 54 60

Lifetime (month)

0

1000

2000

3000

4000

5000

6000

Month

ly T

CO

($)

A, A=100%

B, B

=100%

A=45%,

B=45%

TCOB

<TCOA

6 12 18 24 30 36 42 48 54 60

Server A Lifetime (month)

6

12

18

24

30

36

42

48

54

60

Serv

er

B L

ifetim

e (

month

)

A=45%,

B=77.5%

TCO B<TCO A

TCO A<TCO B

6 12 18 24 30 36 42 48 54 60

Server A Lifetime (month)

6

12

18

24

30

36

42

48

54

60

Serv

er

B L

ifetim

e (

month

)

A=100%,

B=100%

TCO B<TCO A

TCO A<TCO B

6 12 18 24 30 36 42 48 54 60

Server A Lifetime (month)

6

12

18

24

30

36

42

48

54

60

Serv

er

B L

ifetim

e (

month

)

TCO comparison with a Similar-density Server

18

6 12 18 24 30 36 42 48 54 60

Lifetime (month)

400

600

800

1000

1200

1400

Mo

nth

ly T

CO

($

)

A, A=45%

B, B

=45%

6 12 18 24 30 36 42 48 54 60

Lifetime (month)

400

600

800

1000

1200

1400

Mo

nth

ly T

CO

($

)

A, A=45%

B, B

=77.5%

6 12 18 24 30 36 42 48 54 60

Lifetime (month)

0

1000

2000

3000

4000

5000

6000M

on

thly

TC

O (

$)

A, A=100%

B, B

=100%

A=45%,

B=45%

TCOB

<TCOA

6 12 18 24 30 36 42 48 54 60

Server A Lifetime (month)

6

12

18

24

30

36

42

48

54

60

Se

rve

r B

Life

time

(m

on

th)

A=45%,

B=77.5%

TCO B<TCO A

TCO A<TCO B

6 12 18 24 30 36 42 48 54 60

Server A Lifetime (month)

6

12

18

24

30

36

42

48

54

60

Se

rve

r B

Life

time

(m

on

th)

A=100%,

B=100%

TCO B<TCO A

TCO A<TCO B

6 12 18 24 30 36 42 48 54 60

Server A Lifetime (month)

6

12

18

24

30

36

42

48

54

60

Se

rve

r B

Life

time

(m

on

th)

A: Lenovo Flex server B: Our proposed server

𝛿: annual depreciation rate (determines the residual value)

Page 19: Towards Deploying Decommissioned Mobile Devices as Cheap ... · Towards Deploying Decommissioned Mobile Devices as Cheap Energy-Efficient Compute Nodes Mohammad Shahrad and David

Thanks for your attention!

19

Next slide: Discussion Topics

101010

(a) USB Tree A (b) USB Tree B

1010

Master Node Master Node

Smartphones

8,000

80,000

CPU

Mar

k R

atin

g

Release Date 6 12 18 24 30 36 42 48 54 60

Lifetime (month)

400

600

800

1000

1200

1400

Month

ly T

CO

($)

A, A=45%

B, B

=45%

6 12 18 24 30 36 42 48 54 60

Lifetime (month)

400

600

800

1000

1200

1400

Month

ly T

CO

($)

A, A=45%

B, B

=77.5%

6 12 18 24 30 36 42 48 54 60

Lifetime (month)

0

1000

2000

3000

4000

5000

6000

Month

ly T

CO

($)

A, A=100%

B, B

=100%

A=45%,

B=45%

TCOB

<TCOA

6 12 18 24 30 36 42 48 54 60

Server A Lifetime (month)

6

12

18

24

30

36

42

48

54

60

Serv

er

B L

ifetim

e (

month

)

A=45%,

B=77.5%

TCO B<TCO A

TCO A<TCO B

6 12 18 24 30 36 42 48 54 60

Server A Lifetime (month)

6

12

18

24

30

36

42

48

54

60

Serv

er

B L

ifetim

e (

month

)

A=100%,

B=100%

TCO B<TCO A

TCO A<TCO B

6 12 18 24 30 36 42 48 54 60

Server A Lifetime (month)

6

12

18

24

30

36

42

48

54

60

Serv

er

B L

ifetim

e (

month

)

Page 20: Towards Deploying Decommissioned Mobile Devices as Cheap ... · Towards Deploying Decommissioned Mobile Devices as Cheap Energy-Efficient Compute Nodes Mohammad Shahrad and David

Some Discussion Topics

20

•Reliability of used mobile devices under data center workload

•The best management scheme (centralized, distributed, server-level, etc.)

•Using SoC accelerators for domain-specific applications. (security, neural net, etc.)

•Dealing with extreme heterogeneity

•Security concerns with used phones