This is a major change.
-
Work queue structure has been greatly simplified. We got rid of the use of server-side Lua Redis code and programmable task assignment featurs which never worked very well and are no longer supported. Tasks are now consumed by workers with a simple first-come, first-served rule using basic Redis list value types (queues). This simpler scheme should be more robust than the 1.1 versions of the package.
-
Now using the L'Ecuyer RNG functions from the parallel package as the default parallel RNG for reproducible random numbers. User-defined RNGs are still supported as outlined in the vignette.
Set the following parameter in your redis.conf file before using doRedis:
timeout 0
Steve Weston's foreach package is a remarkable parametric evaluation device for the R language. Similarly to lapply-like functions, foreach maps and parameter values expressions to data and aggregates results. Even better, foreach lets you do this in parallel across multiple CPU cores and computers. And even better yet, foreach abstracts the parallel computing details away into modular back-end code. Code written using foreach works sequentially in the absence of a parallel back-end, and works uniformly across a growing variety of back ends. Think of foreach as the lingua Franca of parallel computing for R.
Redis is a powerful, fast networked database with many innovative features, among them a blocking stack-like data structure (Redis "lists"). This feature makes Redis useful as a lightweight backend for parallel computing. The rredis package provides a native R interface to Redis. The doRedis package defines a parallel backend for foreach using Redis that is elastic and platform-independent.
Here is a quick example procedure for experimenting with doRedis:
- Install Redis on your computer.
- Install foreach, rredis and doRedis packages.
- Start the redis server running (see the redis documentation). We assume that the server is running on the host "localhost" and port 6379 (the default Redis port). We assume in the examples below that the worker R processes and the master are running on the same machine. In practice, they can of course run across a network.
- Open one or more R sessions that will act as back-end worker processes. Run the following in each session:
require('doRedis')
redisWorker('jobs')(The R session will display status messages but otherwise block for work.) Note: You can add more workers to a work queue at any time. Also note that each back-end worker may advertise for work on multiple queues simultaneously (see the documentation and examples). 5. Open another R session that will act as the master process. Run the following example (a simple sampling approximation of pi):
require('doRedis')
registerDoRedis('jobs')
foreach(j=1:10,.combine=sum,.multicombine=TRUE) %dopar%
4*sum((runif(1000000)^2 + runif(1000000)^2)<1)/10000000
removeQueue('jobs')Let's define a few terms before we describe how the above example works:
- A loop iteration is the foreach expression together with a single loop parameter value.
- A task is a collection of loop iterations.
- For a given foreach expression, a job is the collection of tasks that make up the full set of loop iterations.
- A work queue is a collection of of any number of tasks associated with number of jobs submitted by one or more master R processes.
The "jobs" parameter above in the redisWorker and registerDoRedis function names a
Redis key used to transfer data between master and worker processes. Think of
this name as a reference to a work queue. The master places tasks into the
queue, worker R processes pull tasks out of the queue and then return their
results to an associated result queue.
The doRedis parallel backend supports dynamic pools of back-end workers. New workers may be added to work queues at any time and can be immediately used by running foreach computations.
The doRedis backend accepts a parameter called chunkSize that sets the number
of loop iterations doled out per task, by default one. Optionally set this with
the setChunkSize function. Increasing chunkSize can improve performance for
quick-running function evaluations. Here is an example that sets chunkSize to
100:
foreach(j=1:5, .options.redis=list(chunkSize=100)) %dopar% ...Setting chunkSize too large will adversely impact load-balancing across
the workers. For instance, setting chunkSize to the total number of loop
iterations will run everything sequentially on one worker!
The redisWorker function is used to manually invoke worker processes that
advertise for job tasks on one or more work queues. The function also has
parameters for a Redis host, port number and password. For example, if the
Redis server is running on a host called "Cazart" with the default Redis port
6379:
redisWorker('jobs', host='Cazart', port=6379)
The registerDoRedis function also contains host and port parameters.
Neither the worker nor master R session needs to be running on the same
machine as the Redis server.
The startLocalWorkers function invokes one or more background R worker
processes on the local machine (internally using the redisWorker function).
It's a convenient way to invoke several workers at once on your local box.
Workers self-terminate when their work queues have been deleted with the
removeQueue function.
[](https://codecov.io/github/bwlewis/doRedis?branch=master)
[](http://cran.rstudio.com/web/packages/doRedis/index.html)