18
1 © 2009-2010 M. Satyanarayanan Microsoft Networking Research Summit, Bellevue, WA, 2010-06-02 The Role of Cloudlets in Mobile Computing Mahadev Satyanarayanan School of Computer Science Carnegie Mellon University

The Role of Cloudlets in Mobile Computing

  • Upload
    liz

  • View
    71

  • Download
    0

Embed Size (px)

DESCRIPTION

The Role of Cloudlets in Mobile Computing. Mahadev Satyanarayanan School of Computer Science Carnegie Mellon University. Machine Translation Today. 0.85. 0.8. Human Scoring Range. 0.7447. 0.7289. 0.7. 0.6. 0.5610. BLEU SCORES. 0.5551. 0.5137. 0.5. 0.4. 0.3859. 0.3. CBMT Spanish. - PowerPoint PPT Presentation

Citation preview

Page 1: The Role of Cloudlets in Mobile Computing

1© 2009-2010 M. Satyanarayanan Microsoft Networking Research Summit, Bellevue, WA, 2010-06-02

The Role of Cloudlets in Mobile Computing

Mahadev Satyanarayanan

School of Computer Science

Carnegie Mellon University

Page 2: The Role of Cloudlets in Mobile Computing

2© 2009-2010 M. Satyanarayanan Microsoft Networking Research Summit, Bellevue, WA, 2010-06-02

Based on same Spanish test set

0.4

0.5

0.6

0.7

0.8

BL

EU

SC

OR

ES

SystranSpanish

GoogleChinese

(‘06 NIST)

CBMTSpanish

0.3GoogleArabic

(‘05 NIST)

0.3859

0.5137

0.5551 0.5610

SDLSpanish

0.85

0.7447

GoogleSpanish

’08 top lang

0.7289

Human Scoring Range

Machine Translation Today

Page 3: The Role of Cloudlets in Mobile Computing

3© 2009-2010 M. Satyanarayanan Microsoft Networking Research Summit, Bellevue, WA, 2010-06-02

Face Recognition Today

Page 4: The Role of Cloudlets in Mobile Computing

4© 2009-2010 M. Satyanarayanan Microsoft Networking Research Summit, Bellevue, WA, 2010-06-02

What’s The Catch?

These are resource-intensive applications

• State-of-art performance and quality only with room full of servers

• How do we achieve this “in the wild”?(on resource-poor, energy-limited mobile hardware)

Obvious solution: leverage the cloud!

But your cloud may be far away

End-to-end latency matters for crisp interaction

• e.g., real-time two-way language translation on mobile devices

• e.g, augmented reality for cognitive assistance via “smart glasses”

and many other examples

Page 5: The Role of Cloudlets in Mobile Computing

5© 2009-2010 M. Satyanarayanan Microsoft Networking Research Summit, Bellevue, WA, 2010-06-02

Latency Hurts Even If Bandwidth Good(E.g. QuakeViz interactive benchmark on VNC thin client 100 Mbps)

Page 6: The Role of Cloudlets in Mobile Computing

6© 2009-2010 M. Satyanarayanan Microsoft Networking Research Summit, Bellevue, WA, 2010-06-02

Sample Internet2 RTTs (milliseconds)

Page 7: The Role of Cloudlets in Mobile Computing

7© 2009-2010 M. Satyanarayanan Microsoft Networking Research Summit, Bellevue, WA, 2010-06-02

Latency on 3G Networks“The wireless delay in the 3G network dominates the whole

network path delay, e.g., latency to the first pingable hop is around

200ms, which is close to the end-to-end Ping latency to landmark

servers distributed across the U.S.”

from “Anatomizing Application Performance Differences on

Smartphones”, to appear in MobiSys 2010 (Huang et al)

Page 8: The Role of Cloudlets in Mobile Computing

8© 2009-2010 M. Satyanarayanan Microsoft Networking Research Summit, Bellevue, WA, 2010-06-02

Solution: Create a Tiny Cloud Nearby

Low-latencyhigh-bandwidth1-hop wireless

network

Olympus Mobile Eye Trek

Wearable Computer

HandtalkWearable

Glove

Nokia N810Tablet

AndroidPhone

Coffee shopCloudlet

WAN todistant cloudon Internet

cloudlet = (compute cluster+ wireless access point+ wired Internet access+ no battery limitations)

““data center in a box”data center in a box”

Page 9: The Role of Cloudlets in Mobile Computing

9© 2009-2010 M. Satyanarayanan Microsoft Networking Research Summit, Bellevue, WA, 2010-06-02

Local Wireless Bandwidth

Original motivation for cloudlets was latency

But 1-hop wireless bandwidth to cloudlet also a win

• wireless LAN bw typically 100X wireless WAN bwe.g. 802.11n ≈ 400 Mbps but HSPDA ≈ 2 Mbps

• shipping large objects within interactive time boundse.g. captured images in an augmented reality system

4MB JPEG image takes 80 ms @ 400 Mbps, but 16 seconds @ 2 Mbps

Page 10: The Role of Cloudlets in Mobile Computing

10© 2009-2010 M. Satyanarayanan Microsoft Networking Research Summit, Bellevue, WA, 2010-06-02

Cloudlet vs. Cloud

CloudletCloudlet CloudCloud

State Only soft state Hard and soft state

Management Appliance model:

self-managed; little

professional attention

Utility model:professionally administered, 24x7 operator coverage

Environment “Data center in a box” at customer premises

Machine room with power conditioning and cooling

Ownership Decentralized ownership by local business

Centralized ownership by Amazon,Yahoo!, etc.

Network LAN latency and bandwidth

Internet latency and bandwidth

Sharing Few users at a time 100s to 1000s of users

Page 11: The Role of Cloudlets in Mobile Computing

11© 2009-2010 M. Satyanarayanan Microsoft Networking Research Summit, Bellevue, WA, 2010-06-02

Key Challenges

1. Trusting infrastructure

• tamper-resistant hardware (“first-world infrastructure”)

• portable device as root of trust (e.g TrustSniffer)

2. Finding the exactly right software on it

uniformity deployer value

specificity end-user valueinherent tension

Page 12: The Role of Cloudlets in Mobile Computing

12© 2009-2010 M. Satyanarayanan Microsoft Networking Research Summit, Bellevue, WA, 2010-06-02

Transient Customization

Deliver fully configured virtual machine (VM) to infrastructure

Problem: too large, too slow for transient use

Solution: assemble VM on the fly dynamic VM synthesis

• prefetch large, relatively static, widely-used piece (“base VM”)

• deliver small patch (“VM overlay”) just before use

• discard VM after use

VM overlay can come from

• mobile device over wireless link, or

• web site under control of mobile device (URL and decryption key)

Page 13: The Role of Cloudlets in Mobile Computing

13© 2009-2010 M. Satyanarayanan Microsoft Networking Research Summit, Bellevue, WA, 2010-06-02

Dynamic VM Synthesis

Mobile

Device

Cloudlet

Preload base VM

Discover & negotiateuse of cloudlet

private VM overlay (base + overlay) launch VM

Execute launch VM

Create VM residue

VM residue

Finish usedone

Depart Discard VM

Optional: cache VM overlay

Usecloudlet

user-drivendevice-VM

interactions

Page 14: The Role of Cloudlets in Mobile Computing

14© 2009-2010 M. Satyanarayanan Microsoft Networking Research Summit, Bellevue, WA, 2010-06-02

Typical Overlay Sizes(base VM = 8GB Ubuntu Linux)

Application Compressed VM Overlay Size (MB)

Uncompressed VM Overlay Size (MB)

Install Package Size (MB)

AbiWord 119.5 364.2 10.0

GIMP 141.0 404.7 16.0

Gnumeric 165.3 519.8 16.0

Kpresenter 149.4 426.8 9.1

PathFind 196.6 437.0 36.8

SnapFind 63.7 222.0 8.8

Null 5.9 24.8 0.0

Page 15: The Role of Cloudlets in Mobile Computing

15© 2009-2010 M. Satyanarayanan Microsoft Networking Research Summit, Bellevue, WA, 2010-06-02

VM Synthesis Time at 100Mbps(untuned proof-of-concept prototype)

0

20

40

60

80

100

120

140

AbiWord GIMP Gnumeric Kpresenter PathFind SnapFind Null

Tim

e in

Sec

on

ds

Other

Resume VM

Apply VM overlay

Decompress VM overlay

Transfer floppy disk

Compress floppy disk

Transfer VM overlay

Largest standard deviation is 5.3% of mean

Nearly half the totalAll in the infrastructurePotentially optimizable

Page 16: The Role of Cloudlets in Mobile Computing

16© 2009-2010 M. Satyanarayanan Microsoft Networking Research Summit, Bellevue, WA, 2010-06-02

When Bandwidth Drops to 10Mbps

0

50

100

150

200

250

300

350

AbiWord GIMP Gnumeric Kpresenter PathFind SnapFind Null

Tim

e in

Sec

on

ds

Other

Resume VM

Apply VM overlay

Decompress VM overlay

Transfer floppy disk

Compress floppy disk

Transfer VM overlay

Largest standard deviation is 3.4% of mean

Page 17: The Role of Cloudlets in Mobile Computing

17© 2009-2010 M. Satyanarayanan Microsoft Networking Research Summit, Bellevue, WA, 2010-06-02

Some Education is Needed

“In the discussion of the proposal, several members of the panel were skeptical about the argument about the need extremely low latencies for handheld devices.

In particular, handhelds are (historically) remarkably powerful computers capable of running user interfaces (the source of most latency sensitivity) locally.

The panel also felt that the case for "cloudlets" was not compelling in contrast to other distributed system architectures such as relying alarge-scale cloud based on conventional data centers and using a geographically distributed three-layer web service architecture.”

Page 18: The Role of Cloudlets in Mobile Computing

18© 2009-2010 M. Satyanarayanan Microsoft Networking Research Summit, Bellevue, WA, 2010-06-02

In Closing

(but keep the Swiss Army Knife handy for emergencies) Leverage the Cloud!Leverage the Cloud!

VM-based CLoudlets