pymemstore

We discovered a problem when writing an app that used multiple processes. We wanted to share a map of data in memory. Sharing it across processes was expensive. We tried using Pipes and Queues. It was just trouble. We were using zeromq on some other projects and decided to give that a try. I wrote a simple master process that holds the map in memory and responds to commands. Then I wrote a client that can talk to it, getting and setting key/values pairs. I also added a set feature that you can have a set of values and ensure their are no duplicates. It worked great. Far easier to setup and ues than the alternatives.

After a bit I realized it needed a way to restore and save it's current state to an from a file. So I added that functionality. Then I thought, wouldn't it be nice to have that state be a file on S3. Boom done.

This is just a start. It's a bit crude. It uses message pack for data storage and message passing. zeromq for communication.

future

Features I'd like to add in the future.

security/authentication - I haven't tested this but I think that other processes could tap into this.
Right now it is tied to using IPC sockets, want to make it support TCP sockets as well.
distributed map - sharding? or just replication to and from other instances.

How to use it

To run the master process.

from pymemstore import MemStore
memstore = MemStore("memstore")
memstore.start()  # starts listening on the ipc socket.

To start with the S3 store support enabled do this instead. (I do want to make this just be part of a settings files instead doing it this way)

from pymemstore.s3store import S3StoredMemStore
memstore = S3StoredMemStore(
    "memstore",
    {
      "AWS_ACCESS_KEY": AWS_ACCESS_KEY,
      "AWS_SECRET_KEY": AWS_SECRET_KEY
    })
memstore.start()  # starts listening on the ipc socket.

Create a Client:

client = MemStoreClient("memstore")
client.start()

Sending commands:

client.send(["<command>", "arg1"..."argN"])

Commands

store

store the current state to a file

>> client.send(["store", "<file location to store state to>"])
["stored", <time in seconds it took>]

restore

restore the current state from a file

>> client.send(["store", "<file location to restore state from>"])
["restored", <time in seconds it took>]

info

Determines the amount of memory being used by the service.

>> client.send(["info"])
[<mb_used>, "mb"]

all

Returns the map of all the data stored. Only really useful for small datasets.

>> client.send(["all", "<table name>"])
{....}

set

Sets a key's value, does not need to be a string, can be any basic datatype.

>> client.send(["set", "<table name>", "<key>", <value>])
1

get

Gets a key's value.

>> client.send(["get", "<table name>", "<key>"])
<value>

done

Shutdown the service

>> client.send(["done"])
["ok"]

push

Pushes a value onto a set. Value must be a hashable type, like a string.

>> client.send(["push", "<set name>", value])
1

pull

Pulls a value off of a set. Value must be a hashable type, like a string.

>> client.send(["pull", "<set name>", value])
1

in_set