CloudNet : Dynamic Pooling of Cloud Resources by Live WAN Migration - - PowerPoint PPT Presentation

▶
cloudnet dynamic pooling of cloud resources by live wan
SMART_READER_LITE
LIVE PREVIEW

CloudNet : Dynamic Pooling of Cloud Resources by Live WAN Migration - - PowerPoint PPT Presentation

Oct 18, 2023 911 likes •1.18k views

CloudNet : Dynamic Pooling of Cloud Resources by Live WAN Migration of Virtual Machines Timothy Wood , Prashant Shenoy University of Massachusetts Amherst K.K. Ramakrishnan, and Jacobus Van der Merwe AT&T Labs - Research VEE 2011

slide-1
SLIDE 1

CloudNet: Dynamic Pooling of Cloud Resources by Live WAN Migration of Virtual Machines

Timothy Wood, Prashant Shenoy University of Massachusetts Amherst K.K. Ramakrishnan, and Jacobus Van der Merwe AT&T Labs - Research VEE 2011

Thursday, March 10, 2011

slide-2
SLIDE 2

Tim Wood - UMass Amherst

Cloud Isolation

  • Cloud data centers are

isolated from one another and the enterprise

  • Complicates
  • Deployment
  • Security
  • Resource management

2

Enterprise View

Need a way to flexibly manage IT resources across data centers

Thursday, March 10, 2011

slide-3
SLIDE 3

Tim Wood - UMass Amherst

  • Flexible cross data center resource pools
  • A secure collection of server, storage, and network resources
  • Seamlessly connected across cloud and enterprise data centers
  • Supports dynamic application placement across sites

Vision: Virtual Cloud Pools

Enterprise Sites Cloud Sites

3

Thursday, March 10, 2011

slide-4
SLIDE 4

Tim Wood - UMass Amherst

Cloud Pool Use Cases

  • Enterprise Consolidation
  • Simplify deployment into the cloud
  • Minimize downtime and reconfiguration
  • Cloud Bursting
  • Many applications cannot be easily replicated
  • WAN Migration enables dynamic placement
  • Minimize performance impact
  • Follow the Sun
  • Application moves to be

closer to clients or data

  • Minimize planning time and

migration bandwidth cost

4

Thursday, March 10, 2011

slide-5
SLIDE 5

Tim Wood - UMass Amherst

Dynamic Cloud Pools

  • Goals:
  • Seamlessly and securely connect enterprise and cloud data centers
  • Enable efficient migration of resources between data centers
  • Challenges
  • Networking: security, transparency, flexibility
  • WAN Migration: Efficiency, application performance impact

5

2 4 6 8 10 12 14 16 250 500 750 1000 Pause Time (sec) Bandwidth (Mbps) SpecJBB Kernel Compile TPC-W

Downtime (sec)

Thursday, March 10, 2011

slide-6
SLIDE 6

Tim Wood - UMass Amherst

Outline

  • Introduction
  • Seamless Connections with VPNs
  • Optimizing WAN Migration of

VMs

  • Implementation & Evaluation
  • Conclusions

6

Thursday, March 10, 2011

slide-7
SLIDE 7

Tim Wood - UMass Amherst

Connectivity Challenges

  • Current approaches lack...
  • Security
  • Firewalls too fine grain, difficult to manage dynamically
  • Transparency
  • Cloud resources have own public IP range separate from enterprise
  • Flexibility
  • Complex reconfiguration required to add resources or move them

between sites

7

Thursday, March 10, 2011

slide-8
SLIDE 8

Tim Wood - UMass Amherst

Seamless Data Center Connections

  • CloudNet: Use

Virtual Private Networks (VPNs)

  • Creates secure end-to-end network paths
  • Simpler configuration than firewalls
  • Managed by network provider with no end host configuration
  • Layer 2

Virtual Private LAN Service (VPLS)

  • Bridges the local networks at multiple sites
  • Makes cloud resources look as if directly attached to

enterprise LAN

  • Allows existing

VM migration techniques to work

  • ver the WAN!

8

Thursday, March 10, 2011

slide-9
SLIDE 9

Tim Wood - UMass Amherst

  • Manipulating

VPN endpoints can be slow

  • Manual process, can take days... must reduce this to seconds
  • CloudNet automates

VPN endpoint reconfiguration

  • Centralized

VPN Controller

  • Acts as route reflector between sites
  • Can adjust ruleset to modify

VPN topology

  • Route updates propagated via BGP

Dynamic VPN Endpoints

9

+ +

VPN Controller

+

Thursday, March 10, 2011

slide-10
SLIDE 10

Tim Wood - UMass Amherst

Outline

  • Introduction
  • Seamless Connections with

VPNs

  • Optimizing WAN Migration of VMs
  • Implementation & Evaluation
  • Conclusions

10

Thursday, March 10, 2011

slide-11
SLIDE 11

Tim Wood - UMass Amherst

WAN Migration Challenges

  • Existing approaches not well optimized for WAN
  • Requires high bandwidth, low latency links (e.g. 622Mbps / 5msec)
  • [VMware/Cisco 09, Travostino 06]
  • Focus only on storage or ignore it completely
  • [Bradford 07, Ruth 06]
  • Need to support moving full

VM state

  • Disk storage
  • Memory data
  • Processor state
  • All with minimal impact on application performance

11

Thursday, March 10, 2011

slide-12
SLIDE 12

Tim Wood - UMass Amherst

VM Migration Procedure

D i s k M e m N e t Asynchronous Copy Synchronous Copy

Live Mem Transfer

Time

ARP VPN Setup Pause VM

12

VM Memory

West Coast East Coast

VM Memory VM Memory VM Memory

Thursday, March 10, 2011

slide-13
SLIDE 13

Tim Wood - UMass Amherst

Optimizing WAN Migration

  • Redundancy Elimination: detect identical regions in

memory or disk and only send once

13

VM Memory

Cache

VM Memory

Cache

Destination Source

Redundancy (% of RAM)

Zeroes

0 10 20 30 40 50 60 70

Non-0 Duplicates

TPC-W Kernel Compile SPECjbb

Thursday, March 10, 2011

slide-14
SLIDE 14

Tim Wood - UMass Amherst

  • Page Deltas: only send delta for partially changed data

blocks during the migration

Deltas

14

Cache

1 2 3 4

2 1

To send

Delta Size (B) Frequency

20K 40K 1000 2000 3000 4000

Destination

1 2

Delta Size (B) Frequency

20K 40K 1000 2000 3000 4000

Page Diff Size Kernel compile TPCW

Thursday, March 10, 2011

slide-15
SLIDE 15

Tim Wood - UMass Amherst

3750 7500 11250 15000 1 2 3 4 5 6 7 8

Number of Pages Iteration

Sent Dirtied

Smart Stop

  • When to stop iterating?
  • Xen: when very few pages

left or after 30 iterations

  • Goals:
  • Minimize total migration time
  • Minimize pause time
  • Iterate until Sent < Dirtied
  • Reduce total time
  • Then, find local minimum for Dirtied
  • Reduce pause time

15

242,987 21,791

Thursday, March 10, 2011

slide-16
SLIDE 16

Tim Wood - UMass Amherst

Outline

  • Introduction
  • Seamless Connections with

VPNs

  • Optimizing WAN Migration of

VMs

  • Implementation & Evaluation
  • Conclusions

16

Thursday, March 10, 2011

slide-17
SLIDE 17

Tim Wood - UMass Amherst

Implementation

  • Uses Xen for memory migration
  • Added in-memory cache for redundancy and deltas
  • Modified control algorithm for smart stop
  • Disk migration based on DRBD
  • Has sync and async modes
  • Evaluated benefits of redundancy elimination with traces
  • VPN controller manages Juniper M7i routers
  • Can be remotely configured
  • Migration wrapper coordinates network, storage, and

memory operations

17

Thursday, March 10, 2011

slide-18
SLIDE 18

Tim Wood - UMass Amherst

CloudNet Testbed

  • Testbed for exploring cloud services
  • 3 sites spread across the US
  • Illinois, Texas, and California
  • Small cluster of servers

at each site

  • Can create multiple

VCPs with resources at each site

  • Migrations performed over

active AT&T network links

18

Thursday, March 10, 2011

slide-19
SLIDE 19

Tim Wood - UMass Amherst

500 mi

Eval: Cloud Burst

Default Optimized

  • From Richardson, TX

to Chicago, IL

  • 465 Mbps link
  • 27 msec RTT
  • >1200 KM distance
  • Simultaneous migration
  • f four

VMs

  • 10GB disk

+ 1.7GB RAM per VM

  • Total BW consumption lowered

from 37GB to 18GB

  • Memory migration time reduced

from 245 to 87 sec

  • Downtime halved from 6 to 3 sec

7 13 20 27 33 40

Memory Disk Total GB Transferred

19

Thursday, March 10, 2011

slide-20
SLIDE 20

Tim Wood - UMass Amherst

  • CloudNet reduces migration time
  • Shorter period with lower application performance
  • Synchronous disk replication reduces application

performance if latency is high

Eval: Application Performance

20

50 100 150 39 40 41 42 43 44 Disk Transfer: 40 min CloudNet: 115s Default Xen Memory Mig: 210s Resp Time (msec) Time (min) Xen CloudNet 38.5

Thursday, March 10, 2011

slide-21
SLIDE 21

Tim Wood - UMass Amherst

Eval: Network Impact

  • Bandwidth: length of migration and pause time
  • Latency: application performance during migration

21

50

100 150 200 250

50

100 1000 Total Time (sec) Bandwidth (Mbps) Xen CloudNet

1 2 3 4 5 6 7 8

50

100 1000 Pause Time (sec) Bandwidth (Mbps) Xen CloudNet

50 100 150 200 0 10 20 30 40 50 60 70 80 Total Time (sec) Latency (msec) Xen CloudNet 1 2 3 4 5 6 0 10 20 30 40 50 60 70 80 Pause Time (sec) Latency (msec) Xen CloudNet

Thursday, March 10, 2011

slide-22
SLIDE 22

Tim Wood - UMass Amherst

Related Work

  • Private Clouds &

Virtual Networks

  • VIOLIN [Ruth-ICAC 06] and

VIRTUOSO [Sundararaj-VM 04]

  • Amazon

VPC

  • Optimizing Migration
  • Compression [Jin-Cluster 09]
  • Model based [Breitgand-HotICE 11]
  • Deltas [Svard-previous talk]
  • Storage transfer [Zheng-next talk]

22

Thursday, March 10, 2011

slide-23
SLIDE 23

Tim Wood - UMass Amherst

Conclusions

  • CloudNet: end-to-end support for WAN migration
  • Network reconfiguration
  • Optimized memory and storage transfer
  • Minimizes migration cost
  • Bandwidth: Eliminate redundant data
  • Time: Reduce unnecessary iterations
  • Reduces application impact
  • Asynchronous bulk disk transfer
  • Minimize pause time

23

Questions?

Thursday, March 10, 2011

slide-24
SLIDE 24

Tim Wood - UMass Amherst

Where to optimize?

  • Can do redundancy elimination within the network
  • Riverbed, [SmartRE]
  • ...but migration data is often encrypted
  • Our end-host based cache can be used for both RE

and page deltas

24

Xen Hypervisor

Dom-0 Virtual Machine

Xen Hypervisor

Dom-0 Virtual Machine

Memory Cache Storage Cache Memory Cache Storage Cache

Thursday, March 10, 2011