jerrinsg / Go Pmem

Introduction

go-pmem is a project that adds native persistent memory support to Go. This is achieved through a combination of language extensions, compiler instrumentation, and runtime changes. Detailed design and implementation details of go-pmem can be found in the ATC 2020 paper on go-pmem. We have also created a website which contains additional documentation and performance reports. go-pmem is based on the 1.15 release version of Go.

How to Build

The persistent memory changes introduced in go-pmem is currently supported only on Linux amd64. To compile the codebase:

$ cd src
$ ./make.bash

The compiled Go binary will be placed in the bin/ directory. An example application (example.go) written to use persistent memory features provided by go-pmem can be found in the design/ folder. To run this application, compile it using the go-pmem binary as shown below:

$ cd design
$ GO111MODULE=off ../bin/go build -txn example.go
$ ./example

The official Go documentation on building the Go compiler can be found here.

Related Projects

go-pmem-transaction

go-pmem-transaction project provides two packages that go-pmem depends on. The pmem package provides APIs to initialize persistent memory and access persistent memory data through named objects. The transaction package provides the transactional functionalities that go-pmem uses for enabling crash-consistent persistent memory data updates. Project home page - https://github.com/vmware/go-pmem-transaction

Go Redis

Go Redis is a Go implementation of Redis designed to run on persistent memory. It is multi-threaded and implements a subset of the Redis commands. The advantages of Go Redis are: much faster database load time on a server restart due to its data being readily available in persistent memory and much higher data throughput thanks to the multithreaded Go implementation. Project home page - https://github.com/vmware-samples/go-redis-pmem