Skip to content

Latest commit

 

History

History
87 lines (57 loc) · 2.57 KB

README.md

File metadata and controls

87 lines (57 loc) · 2.57 KB
Raw

Neks

A dead simple key/value server

Neks is an in-memory networked key/value server written in ~200 lines of Haskell.

It is intended to be very easy to modify.

Features
  • Very simple
  • Pretty fast (>1M transactions per second is easy, see Benchmark)
  • Highly concurrent
  • Atomic transactions (e.g. atomic read-and-swap)
  • Optional disk persistence (with atomic snapshotting)
To install using Cabal:
cabal install neks
NeksServer <opt-args>
NeksClient <args>
To build and run from source (recommended):
cabal sandbox init
cabal install --only-dependencies
cabal run NeksServer
cabal run NeksClient -- <args>

or, with dependencies installed:

ghc -O2 -threaded Network/Neks/NeksServer.hs
ghc -O2 -threaded Network/Neks/NeksClient.hs
./Network/Neks/NeksServer +RTS -N<number of cores>
./Network/Neks/NeksClient <args>

To run the Python client:

python3 Client.py
To view instructions:
NeksServer --help
NeksClient --help
Benchmark:

The server and client run on two cores. The client runs

  • 50 threads
  • 200 requests per thread
  • 50 reads and 50 writes per request

Speed depends on latency and bandwidth. Here's what I get on my home machines:

Avg. Latency Transactions / Second
.1ms (localhost) 1,150,000
.3ms (ethernet) 1,100,000
3.5ms (wireless) 750,000
Protocol:

All network encoding is done using msgpack.

Messages are preceded by the length of the message, transmitted as a 64-bit big-endian unsigned integer.

The client sends requests to the server, and the server responds with the results of the requests.

There are 4 kinds of requests:

  • Requests to Get key K. These are formatted as [0, K]. The server sends a response.
  • Requests to Set key K to value V. These are formatted as [1, K, V]. The server does not send a response.
  • Requests to Delete key K. These are formatted as [2, K]. The server does not send a response.
  • Requests to Atomically evaluate a list of requests R. This is formatted as [3, R]. The server sends the response as if all requests in R had been evaluated normally.

There are 2 kinds of responses:

  • Response that the value V for requested key K was found. This is formatted as [-1, V].
  • Response that the value for requested key K was not found. This is formatted as [-2].

Example conversation:

Client: <message length>[[1,"status","OK"],[0,"Jim"],[0,"Dwight"]]
Server: <message length>[[-1,"Halpert"],[-1,"Schrute"]]