FairCloud: Sharing the Network in Cloud Computing Presenter: Wei sun Lucian popa Gautam Kumar Mosharaf Chowdhury (HP Labs (UC Berkeley (UC Berkeley Arvind Krishnamurthy Sylvia Ratnasamy lon stoica (Univ Washington) (UC Berkeley (UC Berkeley
FairCloud: Sharing the Network in Cloud Computing Lucian Popa Gautam Kumar Mosharaf Chowdhury Arvind Krishnamurthy Sylvia Ratnasamy Ion Stoica (UC Berkeley) (UC Berkeley) (Univ Washington) (UC Berkeley) (UC Berkeley) (HP Labs) Presenter: Wei Sun
Motivation Request Instances Wizard Cancel x CH005E具NAM INSTANCE DETAILS CLATEKEY PA Provide the details for your instance(s). You may also decide whether you want to launch your instances as"on-demand"or Number of Instances: 1 Instance Type: Micro(t1micro, 6/ MB) Type CPU Units CPU res Memory o Launch I Micro(tI micro) Up to 2 ECUs 1 Core 613MB EC2 Instance ms what are commonly lard Large(m1.large) 4 ECUs 7.5GB Launch into Extra Large(m1xlarge) 15 GB High- Memory Double Extra Large(m2. 2xlarge) 13 ECUs cores O Request High-Memory Quadruple Extra Large(m2. 4xlarge\ 26 ECUs 8 Cores/ 68.4 GE High-CPU Extra Large(c1. xlar ECUs 7 GB Network?
Motivation Network?
Context Networks are more difficult to share than other resources
Context Networks are more difficult to share than other resources X
Context Several proposals that share network differently eg proportional to source VMs Seawall [NSDl11]) statically reserve bandwidth(oktopus [ Sigcomm12]) Provide specific types of sharing policies Characterize solution space and relate policies to each other
Context • Several proposals that share network differently, e.g.: – proportional to # source VMs (Seawall [NSDI11]) – statically reserve bandwidth (Oktopus[Sigcomm12]) – … • Provide specific types of sharing policies • Characterize solution space and relate policies to each other?
This talk → 1. Framework for understanding network sharing in cloud computing Goals, tradeoffs, properties 2. Solutions for sharing the network Existing policies in this framework New policies representing different points in the design space
This Talk 1. Framework for understanding network sharing in cloud computing – Goals, tradeoffs, properties 2. Solutions for sharing the network – Existing policies in this framework – New policies representing different points in the design space
Goals 1. Minimum Bandwidth guarantees Provides predictable performance Example: file transfer finishes within time limit min A 1 A Timemax =Size/bmin
Goals 1. Minimum Bandwidth Guarantees – Provides predictable performance – Example: file transfer finishes within time limit A1 A2 Timemax = Size / Bmin Bmin
Goals 1. Minimum Bandwidth guarantees 2. High Utilization Do not leave useful resources unutilized Requires both work-conservation and proper Incentives AB B B Both tenants active Non work-conserving Work-conserving
Goals 1. Minimum Bandwidth Guarantees 2. High Utilization – Do not leave useful resources unutilized – Requires both work-conservation and proper incentives A B B B Both tenants active Non work-conserving Work-conserving
Goals 1. Minimum Bandwidth guarantees 2. High Utilization 3. Network Proportionality As with other services, network should be shared proportional to payment Currently tenants pay a flat rate per Vmi network share should be proportional to #VMs assuming identical VMs)
Goals 1. Minimum Bandwidth Guarantees 2. High Utilization 3. Network Proportionality – As with other services, network should be shared proportional to payment – Currently, tenants pay a flat rate per VM ➔ network share should be proportional to #VMs (assuming identical VMs)
Goals 1. Minimum Bandwidth guarantees 2. High Utilization 3. Network Proportionality Example: a has 2 VMs, b has 3 VMs BW△ BWB(B3 BW 2 BWB 3 B When exact sharing is not B possible use max-min
Goals 1. Minimum Bandwidth Guarantees 2. High Utilization 3. Network Proportionality – Example: A has 2 VMs, B has 3 VMs A1 A2 BwA B1 B3 B2 BwB BwB BwA = 2 3 When exact sharing is not possible use max-min