Sharing Cloud Networks Lucian Popa, Gautam Kumar, Mosharaf Chowdhury Arvind Krishnamurthy, Sylvia Ratnasamy, Ion Stoica UC Berkeley

State of the Cloud Network??? 2

Guess the Share 3 B2B2 A1A1 B1B1 A2A2 A3A3 Link L Alice : Bob = ? : ? 3 : 11 : 12 : 13 : 2 TCP Per flowPer Source Per Destination Per-VM Proportional

Challenges Network share of a virtual machine (VM) V depends on »Collocated VMs, »Placement of destination VMs, and »Cross-traffic on each link used by V Network differs from CPU or RAM »Distributed resource »Usage attribution (source, destination, or both?) Traditional link sharing concepts needs rethinking 4

Requirements 5 High Utilization Min Bandwidth Guarantee Aggregate Proportionality Network shares proportional to the number of VMs Introduce performance predictability Do not leave bandwidth unused if there is demand

Requirement 1: Guaranteed Minimum B/W Provides a minimum b/w guarantee for each VM Captures the desire of tenants to get performance isolation for their applications 6 ABX B minA B minB B minX … All VMs

Requirement 2: Aggregate Proportionality Shares network resources across tenants in proportion to the number of their VMs Captures payment-proportionality »Similar to other resources like CPU, RAM etc. Desirable properties »Strategy-proofness: Allocations cannot be gamed »Symmetry: Reversing directions of flows does not change allocation 7

Design Space 8 High Utilization Min Bandwidth Guarantee Aggregate Proportionality Strategy- Proofness Symmetry

Requirement 3: High Utilization Provides incentives such that throughput is only constrained by the network capacity »Not by the inefficiency of the allocation or by disincentivizing users to send traffic Desirable properties »Work Conservation: Full utilization of bottleneck links »Independence: Independent allocation of one VM’s traffic across independent paths 9

Tradeoff 1 Design Space 10 High Utilization Min Bandwidth Guarantee Aggregate Proportionality Work Conservation Strategy- Proofness IndependenceSymmetry

Tradeoff 1: Min B/W vs. Proportionality 11 Share of Tenant A can decrease arbitrarily! B2 A1 B1 A2 Link L with Capacity C C/2 B3 C/3 2C/3

Tradeoff 1 Tradeoff 2 Design Space 12 High Utilization Min Bandwidth Guarantee Aggregate Proportionality Work Conservation Strategy- Proofness IndependenceSymmetry

Tradeoff 2: Proportionality vs. Utilization 13 A1A1 A3A3 B1B1 B3B3 A2A2 A4A4 B2B2 B4B4 Link L with Capacity C C/4 To maintain proportionality, equal amount of traffic must be moved from A1-A2 to A1-A3 => Underutilization of A1-A3

Per-link Proportionality Restrict to congested links only Share of a tenant on a congested link is proportional to the number of its VMs sending traffic on that link 14

Per Endpoint Sharing (PES) Five identical VMs (with unit weights) sharing a Link L 15 B2B2 A1A1 B1B1 A2A2 A3A3 Link L

Per Endpoint Sharing (PES) Resulting weights of the three flows: 16 Link L 3/2 2 To generalize, weight of a flow A-B on link L is.. WAWA W A-B = NANA NBNB WBWB + N A = 2 B2B2 A1A1 B1B1 A2A2 A3A3

Per Endpoint Sharing (PES) Symmetric Proportional »sum of weights of flows of a tenant on a link L = sum of weights of its VMs communicating on that link Work Conserving Independent Strategy-proof on congested links 17 WAWA W A-B = NANA NBNB WBWB + = W B-A

Generalized PES 18 AB α > β if L is more important to A than to B (e.g., A’s access link) WAWA W A-B = W B-A = α NANA NBNB WBWB + β+ β L Scale weight of A by α Scale weight of B by β

One-Sided PES (OSPES) 19 AB α = 1, β = 0 if A is closer to L α = 0, β = 1 if B is closer to L WAWA W A-B = W B-A = α NANA NBNB WBWB + β+ β L Scale weight of A by α Scale weight of B by β Per Source closer to source Per Destination closer to destination Highest B/W Guarantee *In the Hose Model

Comparison 20 Per Flow Per Source Static Reservation Link PESOSPES Link Proportionality   Symmetry ✓✗✓✓✓  Strategy-Proofness ✗✓✓✓✓ Utilization   Independence ✓✓✓✓✓  Work Conservation ✓✓✗✓✓ B/W Guarantee 

Full-bisection B/W Network 21 Tenant 1 has one-to-one communication pattern Tenant 2 has all-to-all communication pattern OSPES

MapReduce Workload 22 W1:W2:W3:W4:W5 = 1:2:3:4:5 OSPES

Summary Sharing cloud networks is all about making tradeoffs »Min b/w guarantee VS Proportionality »Proportionality VS Utilization Desired solution is not obvious »Depends on several conflicting requirements and properties »Influenced by the end goal 23