DonDon t’t Lose Sleep Over Availability - USENIX · DonDon t’t Lose Sleep Over Availability:...

Don’t Lose Sleep Over Availability:Don t Lose Sleep Over Availability: The GreenUp Decentralized Wakeup Service

Siddhartha Sen, Princeton University

J b R L h Ri h d H h C l G J SJacob R. Lorch, Richard Hughes, Carlos G. J. Suarez, Brian Zill, Weverton Cordeiro, and Jitendra Padhye

Enterprise networksEnterprise networks

users, IT admins

Despite the cloud, this is a common scenario

Energy savings

Stay Green!

Availability

Stay Up!

Energy savings Availability

Energy savings

Stay Green!

Availability

Stay Up!

GreenUp!Energy savings AvailabilityGreenUp!

Sleep proxySleep proxy

Machine(active)

Machine(active) Sleep proxy

Machine(asleep) Sleep proxy

Send traffic to me!Machine(asleep) Sleep proxy

Send traffic to me!

Machine(asleep) Sleep proxy

Machine(asleep) Machine

(asleep)Sleep proxy

Remote requestq(TCP SYN)

(asleep)Sleep proxy

(asleep)Sleep proxyRemote request

(TCP SYN)

Wake up!(WoL)Machine

(asleep) Machine(asleep)

Sleep proxy(WoL)

(asleep)Sleep proxyWake up!

Remote requestq(TCP SYN)

Machine(active) Sleep proxyRemote request

(TCP SYN)

Response

RResponse

Pros Cons

WAN• No special hardware• No envir. changesN h

• Dedicated server per subnet

• No app changes

Dedicated servers are a problemDedicated servers are a problem

• High deployment andHigh deployment and management cost

• Single point of failure

Dedicated servers are a problemDedicated servers are a problem

• High deployment andHigh deployment and management cost

• Single point of failure

• High availability becomes expensive!

GreenUp A decentralized minimalGreenUp: A decentralized, minimal software‐only sleep proxy

Any machine can act as a proxy (manager) forAny machine can act as a proxy (manager) for sleeping machines on the subnet

OutlineOutline1. How does GreenUp work?2. What can I learn from GreenUp?

3. How effective is GreenUp?– Evaluation on 100 user machines, currently y

deployed on 1,100machines

Machine State… …… …… …

Distributed management

Subnet state coordination

Guardians

GreenUp’s environmentGreenUp s environment

• Subnet domainsSubnet domains

• Load‐sensitive, unreliable machinesLoad sensitive, unreliable machines

• Single administrative domainSingle administrative domain

• Availability most importanty p

Running example (not to scale)Running example (not to scale)

M6awake

Running example (not to scale)Running example (not to scale)

M6awake

asleep + unmanaged

Running example (not to scale)Running example (not to scale)manager

M1 M8 asleep + managedM5

M2 awake

managedM5

M9M4asleep +

unmanaged

Distributed management: hWho manages M9?

b f l– No guarantees before sleep– M1 failure abandons M8M1 M8

M5• Probe randomly, repeat since machines unreliable

• Load‐sensitive machines, di ib bi

M7 so distribute probing– Robust to manager issuesM4

• Wait for notification?– No guarantees before sleepM1 f il b d M8

M5 – M1 failure abandons M8

• Probe randomly, repeatM2

M6 Probe randomly, repeat since machines unreliableM3

M7• Load‐sensitive machines, so distribute probing

– Robust to manager issues

M1 M8total #

machines

machinesawake#nM2

M6 machinesawake#M3

M1 M8total #

machines# managed

M6 machinesawake#imn M2

machinesawake#M3

M1 M8total #

machines# managed

M6 machinesawake#imn M2

machinesawake#M3

machinesawake#ln 1

1pimn M2

M7 • Coupon collector analysis

machinesawake#M3

M7 Coupon collector analysisM4

machinesawake#ln 1

1pimn M2

machinesawake#M3

M1 M8p = Pr(machine probed)

machinesawake#ln 1

1pimn M2

machinesawake#M3

M1 M8p = Pr(machine probed)

machinesawake#ln 1

1pimn M2

machinesawake#M3

Multiple managersMultiple managers

M2 machinesawake#ln 1

M7 • Availability most importantM3

machinesawake#M6

M7 Availability most important

• Simple resolution protocolM4

machinesawake#M6M9

machinesawake#M6

Load balanceLoad balance

M8 M9M1

M6M3 M9

M7M4 M1

• Induction analysis: equivalent to balls‐in‐bins!

M6)2/ln(ln

)2/ln(nn after n/2 sleeps

M3 M9M6

M7M4 M1

• Induction analysis: equivalent to balls‐in‐bins!

Distributed management elects leaders in aM2 M8

M6)2/ln(ln

)2/ln(nn after n/2 sleeps

Distributed management elects leaders in a robust and load‐balanced way, assuming temporary conflicts are tolerable

M3 M9M6

temporary conflicts are tolerable.

M4 M1M7

Subnet state coordinationSubnet state coordination

• Distributed management relies on global state

M5 g– Who to probe?– How to manage?

• IP address, MAC addressM3

M7• TCP listen ports

• Distributed management relies on global state

M5 g– Who to probe?– How to manage?

M2Machine StateM8 …M9

• IP address, MAC addressM3M9 …M5 …… …

M7• TCP listen ports

• Replicated state machine?– Unreliable machines, correlated behavior

M5 correlated behavior– Strong consistency overkillM2Machine State

M8 …M9

M6• External database?– Lose instant deployability

M3M9 …M5 …… …

• Exploit subnet and weaker M9M4

consistency

M8 …M9

M3M9 …M5 …… …

consistency

M8 …M9

M3M9 …M5 …… …

consistency

1. Periodic broadcast while awake

2. Rebroadcast by h l l

M6 managers while asleep

3 Daily roll call toM3

3. Daily roll call to garbage‐collect stateM4

M9 state

3. Daily roll call to garbage‐collect state

M6 ? managers while asleep

M1 M8M8 state

M6M8’ state

managers while asleep

M8’ state

M5Subnet state coordination distributes2. Rebroadcast by

Subnet state coordination distributes per‐machine state on a subnet when strong consistency is not requiredmanagers while asleep

strong consistency is not required.

Guardians

P iMachine State… …… …… …

• Protects against simultaneous sleep

• Caps the max loadDistributed management

Guardians• Caps the max load

Guardians

Deployment in MicrosoftDeployment in Microsoft

• C# codeC# code– Interfaces with packet sniffer/network driver

• Client GUI for users and d l teasy deployment

• Pilot on 1 100 machinesPilot on 1,100 machines

EvaluationEvaluation

• Logs from 101 Windows 7 machines Feb –Logs from 101 Windows 7 machines, Feb. Sep. 2011

• Questions:– Does GreenUp consistently wake machines whenDoes GreenUp consistently wake machines when accessed?

– Does it do so in time to meet user patience?Does it do so in time to meet user patience?– Can GreenUp scale to large subnets?

GreenUp wakes machines reliablyGreenUp wakes machines reliably

• Connect to machines usingConnect to machines using Samba (TCP port 139)

• 11 different days (weekends, evenings):– 496 already awake, 278 woken, 5 unwakeable

– Most failures due to WoL– Most failures due to WoL

• 99.4% success rate

GreenUp wakes machines reliablyGreenUp wakes machines reliably

• Connect to machines usingConnect to machines using Samba (TCP port 139)

• 11 different days (weekends, evenings):

WoL is availability bottleneck!

– 496 already awake, 278 woken, 5 unwakeable

– Most failures due to WoL

bott e ec

– Most failures due to WoL

• 99.4% success rate

GreenUp wakes machines quicklyGreenUp wakes machines quickly

• GreenUp relies on someGreenUp relies on some user patience– Wakeup delay– User retry logic

• Side‐effect of WoLfailure: manager logs how long user waitshow long user waits– 48 events

• GreenUp relies on someGreenUp relies on some user patience– Wakeup delay– User retry logic

• Side‐effect of WoLfailure: manager logs how long user waitshow long user waits– 48 events

• Convolving: GreenUp wakes machines before userConvolving: GreenUp wakes machines before user gives up 85% of the time

87% of wakeups take < 9 sec

13% of users give up after 3 sec (port scanners?)

GreenUp scales to large subnetsGreenUp scales to large subnets

• Sources of manager loadSources of manager load– Intercept traffic for asleep machinesBroadcast state– Broadcast state

– Probe/respond to probes

Good load balance +h k henough awake machines

few managed machines!

• Simulated probing load on 2.4‐GHz, dual‐coreSimulated probing load on 2.4 GHz, dual core Windows 7 machine w/ 4GB memory and 1Gb/s NIC:

# of managed machines CPU utilization100 12%200 21%300 29%

• Guardians ensure max load is 100

• Simulated probing load on 2.4‐GHz, dual‐coreSimulated probing load on 2.4 GHz, dual core Windows 7 machine w/ 4GB memory and 1Gb/s NIC:

# of managed machines CPU utilization100 12%200 21%300 29%

• Guardians ensure max load is 100

Does GreenUp save more energy?Does GreenUp save more energy?

• Energy savings depends on sleep timeEnergy savings depends on sleep time

Average 31% $17.50/machine/year

• IT enforces sleep policy at Microsoft, so hardIT enforces sleep policy at Microsoft, so hard to tell

Extension: Higher availability via l l d h d ffexplicit load hand‐off

M8 M9M1

M7M8 M9M1

• Theorem Expected maxM2

M6 • Theorem. Expected max load = n/d Hd.M3

# awake machines

Harmonic numbers

M7M8 M9M1

DonDon t’t Lose Sleep Over Availability - USENIX · DonDon t’t Lose Sleep Over Availability:...

Documents

Transcript of DonDon t’t Lose Sleep Over Availability - USENIX · DonDon t’t Lose Sleep Over Availability:...

Biology m5

R..M5 M5 M5 M5 M5 M5 M2 M2 M2 M2 M2 F.. W.. S.. K.. R.. 5 General information Mounting Positions 70 Gear Units – 2016 Symbols The following table shows the symbols used in th e mounting

M5 Tutorial

COMMUNICABLE DISEASES GUIDELINES FOR FIREFIGHTERS · SOP: Suspicious Letter • DonDon t’t shake or open shake or open • Place in a plastic bag or other container ... – MRSA

M5 - DYNAMIQUE

m5 Heuristics

PELICULAS M5

Diastrophism m5

Anviz M5 Quick Guide-M5 V1.4-F.02.02185-161013

M5 CONCLUSIONES

Meeting 1 M5 Link Tunnels and New M5 M4 - Westconnex

The Unpaid Intern: CanCan t’t Live - PMBA | Public … Unpaid Intern: CanCan t’t Live With Them, But Can’t Live Without Them Nancy Cooper and JdithJudith EdjEndejan Garvey Schubert

M5 socializacion

delwp-internet-m5.clients.squiz.net€¦ · Web viewdelwp-internet-m5.clients.squiz.net

CARP REMOVAL AT RIVER MURRAY LOCK AND WEIRSM_s3_B3_Brenton+Erd… · carp jump when they meet a barrier but native fish dondon t’t. • He built the first carp separation cage based

SUUNTO M5 GebruikershandleidinGns.suunto.com/Manuals/M5/Userguides/Suunto_M5_UserGuide_NL.pdf2 WELKOM Dank u voor het kiezen van de Suunto M5 hartslagmonitor! De Suunto M5 hartslagmonitor

M M5-091 M5-111 - New Holland Rochester...M M5-091 KUBOTA DIESEL TRACTOR/M5-111 A new wide cabin, dramatically cleaner emissions, and powerful engines highlight the new M5-091/M5-111

!#$%&' gwl6-m5 gws6-m5 sc3w-m5-4 sc3w-m5-6 gws4-m5-s gws4-m5 gwl4-m5 gwl6-m5 sc3w-6-4・6・8 gws4-6 gws6-6 15gws8-6 gwl4-610 gwl6-6 φ12以下 ...

Transcript M5

Agency FleetFocus M5 Instructions COMMUTE USAGE ENTRY Forms/M5... · FleetFocus M5 (M5) is a comprehensive vehicle management system used to track and manage the Commonwealth Fleet.