BwE

Abstract

WAN bandwidth remains a constrained resource that is economically infeasible to substantially overprovision. Hence, it is important to allocate capacity according to service priority and based on the incremental value of additional allocation. For example, it may be the highest priority for one service to receive Gb/s of bandwidth but upon reaching such an allocation, incremental priority may drop sharply favoring allocation to other services. Motivated by the observation that individual ows with xed priority may not be the ideal basis for bandwidth allocation, we present the design and implementation of Bandwidth Enforcer (BwE), a global, hierarchical bandwidth allocation infrastructure. BwE supports: i) service-level bandwidth allocation following prioritized bandwidth functions where a service can represent an arbitrary collection of ows, ii) independent allocation and delegation policies according to user-de ned hierarchy, all accounting for a global view of bandwidth and failure conditions, iii) multi-path forwarding common in tra cengineered networks, and iv) a central administrative point to override (perhaps faulty) policy during exceptional conditions. BwE has delivered more service-e cient bandwidth utilization and simpler management in production for multiple years.

Rethinking networking abstractions for cloud tenants

“…There is a significant amount of prior work in distributed rate limiting within a datacenter [5,13] and traffic engineering across cloud WANs [11,12,14] that suggest that our QoS proposal is feasible. However, our QoS API depends on the combination of these two areas: enforcing bandwidth guarantees per tenant beyond the scope of a singular datacenter.…”

Congestion Control for Cross-Datacenter Networks

“…Specifically, running legacy reactive transport protocols over shallow-buffered WAN may lead to either high packet losses or degraded throughput or both. To mitigate these problems, current practice seeks help from global traffic engineering [64,55,56], endhost rate limiting [71], or traffic scheduling with differentiated services.…”

“…It distributes network traffic to multiple site-to-site paths (usually hundreds of updates per day). Secondly, bandwidth allocation [71] applies to the tenant or task flow group level. It re-allocates the site-to-site bandwidths and split them among all competing flow groups.…”

“…With the prevalence of the geo-distributed applications and services, inter-datacenter network (IDN) is becoming an increasingly important cloud infrastructure [64,55,71,56].…”

IntSight

“…Active measurement techniques (e.g., ping, traceroute, OWAMP [26], and TWAMP [17]) also do not provide sufficient time resolution; additionally, there is no guarantee that the network will route and prioritize probes in the same way as production packets. The consistent move towards heterogeneity in the treatment of traffic [18,20], multi-path routing [19,22], and flowlet load balancing [1,16] exacerbates this limitation.…”