Abstract
Nowadays, most applications hosted on public cloud data centers (DCs) disseminate data from a single source to a group of receivers for service deployment, data replication, software upgrade, etc. For such one-to-many data communication paradigm, multicast routing is the natural choice as it reduces network traffic and improves application throughput. Unfortunately, recent approaches adopting IP multicast routing suffer from scalability and load balancing issues, and do not scale well with the number of supported multicast groups when used for cloud DC networks. Furthermore, IP multicast does not exploit the topological properties of DCs, such as the presence of multiple parallel paths between end hosts. Despite the recent efforts aimed at addressing these challenges, there is still a need for multicast routing protocol designs that are both scalable and load-balancing aware. This paper proposes Ernie, a scalable load-balanced multicast source routing for large-scale DCs. At its heart, Ernie further exploits DC network structural properties and switch programmability capabilities to encode and organize multicast group information inside packets in a way that minimizes downstream header sizes significantly, thereby reducing overall network traffic. Additionally, Ernie introduces an efficient load balancing strategy, where multicast traffic is adequately distributed at downstream layers. To study the effectiveness of Ernie, we extensively evaluate Ernie’s scalability behavior (i.e., switch memory, packet size overheads, and CPU overheads), and load balancing ability through a mix of simulation and analysis of its performances. For example, experiments of large-scale DCs with 27k+ servers show that Ernie requires a downstream header sizes that are 10× smaller than those needed under state-of-the-art schemes while keeping end-host overheads at low levels. Our simulation results also indicate that at highly congested links, Ernie can achieve a better multicast load balancing than other existing schemes.
Highlights
The past decade has witnessed a rapid boom of cloud computing services
We introduce an effective multicast traffic load balancing technique for downstream links that assigns multicast groups to core switches in a way that ensures the evenness of load distribution across the downstream links
In multi-tenant datacenters with 27k servers, our experiments show that the proposed techniques require a downstream header size that is 10× and 3× smaller than that needed under Elmo and Bert, respectively, and achieve between 25 and 65% load balancing improvement over other schemes for highly congested network links
Summary
The past decade has witnessed a rapid boom of cloud computing services. Large cloud service providers, such as Amazon AWS [1], Microsoft Azure [2] and Google Cloud Platform [3], host hundreds of thousands of tenants, each of which possibly running hundreds of applications. Proposed source-routed multicast schemes for cloud DCs, such as Elmo [15] and Bert [16], address the scalability limitations and are shown to scale well with millions of multicast groups They do so by exploiting both the DC network topology symmetry and hardware switch programmability to efficiently encode multicast routing information inside packets. Bert [16] aims to alleviate network overhead incurred by Elmo, it imposes bandwidth and end-host CPU overheads when the number of clusters (packet replications at the source) is large These schemes neglected the multicast traffic load balancing and relied on underlying multipathing protocols (e.g., ECMP [17]). Ernie reconsiders possible solutions by further leveraging the topological properties of modern DC architectures It scales to much larger numbers of multicast groups, while minimizing network overhead (i.e., switch memory and packet size overheads) and with an eye towards downlinks loads (highly congested links).
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
