Datacenter RPCs can be General and Fast

06/02/2018
by   Anuj Kalia, et al.
0

It is commonly believed that datacenter networking software must sacrifice generality to attain high performance. The popularity of specialized distributed systems designed specifically for niche technologies such as RDMA, lossless networks, FPGAs, and programmable switches testifies to this belief. In this paper, we show that such specialization is not necessary. eRPC is a new general-purpose remote procedure call (RPC) library that offers performance comparable to specialized systems, while running on commodity CPUs in traditional datacenter networks based on either lossy Ethernet or lossless fabrics. eRPC performs well in three key metrics: message rate for small messages; bandwidth for large messages; and scalability to a large number of nodes and CPU cores. It handles packet loss, congestion, and background request execution. In microbenchmarks, one CPU core can handle up to 10 million small RPCs per second, or send large messages at 75 Gbps. We port a production-grade implementation of Raft state machine replication to eRPC without modifying the core Raft source code. We achieve 5.5 microseconds of replication latency on lossy Ethernet, which is faster than or comparable to specialized replication systems that use programmable switches, FPGAs, or RDMA.

READ FULL TEXT

page 1

page 2

page 3

page 4

research
10/23/2020

The nanoPU: Redesigning the CPU-Network Interface to Minimize RPC Tail Latency

The nanoPU is a new networking-optimized CPU designed to minimize tail l...
research
05/25/2020

TeaMPI – Replication-based Resilience without the (Performance) Pain

In an era where we can not afford to checkpoint frequently, replication ...
research
03/11/2020

Constellation: A High Performance Geo-Distributed Middlebox Framework

Middleboxes are increasingly deployed across geographically distributed ...
research
06/20/2022

Building Blocks for Network-Accelerated Distributed File Systems

High-performance clusters and datacenters pose increasingly demanding re...
research
01/20/2022

The Specialized High-Performance Network on Anton 3

Molecular dynamics (MD) simulation, a computationally intensive method t...
research
07/01/2019

Creek: a General Mixed-Consistency Transactional Replication Scheme

In this paper we introduce Creek, a low-latency, eventually consistent r...
research
10/05/2021

SiliFuzz: Fuzzing CPUs by proxy

CPUs are becoming more complex with every generation, at both the logica...

Please sign up or login with your details

Forgot password? Click here to reset