Techniques for Memory-efficiency, Low-latency and High-throughput in RDMA-based Datacenter Networks and Applications

Event Sponsor: 
Mathematics and Computer Science Division Seminar
Start Date: 
Jun 12 2017 - 10:30am
Building/Room: 
Building 240/Room 1406-1407
Location: 
Argonne National Laboratory
Speaker(s): 
Jiachen Xue
Host: 
Kamil Iskra

Remote Direct Memory Access (RDMA) fabrics such as Infiniband and Converged Ethernet report latencies that is shortened by a factor of 50 as compared to TCP. Therefore, RDMA augurs well for the emerging class of user-facing, low-latency applications, such as Web search and Memcached, and is a potential replacement for TCP in datacenters (DCs). In this talk, I will mainly cover two aspects of RDMA based distributed system design: (1) RIMA for efficient RDMA buffer management and (2) a client-centric RDMA based Memcached design.
 
Existing RDMA buffer management scheme either incurs memory wastage or involves significant programming effort. In contrast, my proposal, RIMA (remote indirect memory access), avoids the above pitfalls by using indirection. The use of indirection enables RIMA to employ 'append' semantics which is vastly more efficient than existing RDMA communication semantics as it eliminates the need for over-allocation and requires no additional programming effort.
 
I will also present a set of techniques to maximize throughput and to minimize latency for an important datacenter workload -- Memcached. In addition to improving performance at the RDMA communication layer, the proposed techniques also help improve end-to-end latency and throughput. Existing Memcached designs suffer from performance bottlenecks because they focus on optimizing server-side throughput. My design optimizes client-side throughput which is more valuable as it directly impacts the front-end servers that access storage and memory caching tiers. The resulting Memcached implementation achieves significantly better throughput and latency than recently proposed RDMA-based key-value stores.
 
Speaker Biography:
Jiachen Xue recently received his PhD degree in electrical and computer engineering from Purdue university, West Lafayette. His research interests are high performance RDMA based network for distributed system, software/hardware co-design and computer architecture. He received his B.S degree in software engineering from Beihang University, Beijing, China; and M.S. degree in electrical engineering from Arizona State University, Tempe.