Package 'RedisParam'

Deprecated functions in the RedisParam package

Description

bpstopall() is provided for compatibility with previous versions of RedisParam, and will be defunct after the next release. Use rpstopall() instead.

Usage

bpstopall(x)

Arguments

x	a RedisParam object.

Value

See ?rpstopall for return value.

Examples

if (FALSE) {
    ## bpstopall()
    ## deprecated -- use rpstopall() instead
}

---

Enable redis-based parallel evaluation in BiocParallel

Description

RedisParam() creates an object describing manager and worker configurations for parallel compuation using a Redis server back-end.

rpalive() tests whether it is possible to connect to a redis server using the host, port, and password in the RedisParam object.

rpstopall() is used from the manager to stop redis workers launched independently, with is.worker=TRUE.

rpworkers() determines the number of workers using snowWorkers() if workers are created dynamically, or a fixed maximum (currently 1000) if workers are listening on a queue.

rphost() reads the host name of the Redis server from the system environment variable REDISPARAM_HOST, if the variable is not defined, fallback to REDIS_HOST. Otherwise default to "127.0.0.1". rphost(x) gives the host name used by x.

rpport() reads the port of the Redis server from a system environment variable REDISPARAM_PORT, if the variable is not defined, fallback to REDIS_PORT. Otherwise default to 6379. rpport(x) gives the port used by x.

rppassword() reads an (optional) password from the system environment variable REDISPARAM_PASSWORD, if the variable is not defined, fallback to REDIS_PASSWORD. Otherwise default to NA_character_ (no password). rppassword(x) gives the password used by x.

Usage

RedisParam(
  workers = rpworkers(is.worker),
  tasks = 0L,
  jobname = ipcid(),
  log = FALSE,
  logdir = NA,
  threshold = "INFO",
  resultdir = NA_character_,
  stop.on.error = TRUE,
  timeout = NA_integer_,
  exportglobals = TRUE,
  progressbar = FALSE,
  RNGseed = NULL,
  queue.multiplier = 2L,
  redis.hostname = rphost(),
  redis.port = rpport(),
  redis.password = rppassword(),
  is.worker = NA
)

rpalive(x)

rpstopall(x)

rpworkers(is.worker)

rphost(x)

rpport(x)

rppassword(x)

rpisworker(x)

## S4 method for signature 'RedisParam'
bpisup(x)

## S4 method for signature 'RedisParam'
bpbackend(x)

## S4 method for signature 'RedisParam'
bpstart(x, ...)

## S4 method for signature 'RedisParam'
bpstop(x)

## S4 method for signature 'RedisParam'
bpworkers(x)

## S4 replacement method for signature 'RedisParam,logical'
bplog(x) <- value

Arguments

workers	integer(1) number of redis workers. For is.worker=FALSE, this parameter is the maximum number of workers expected to be available. For is.worker=NA, this is the number of workers opened by bpstart().
tasks	See ?"BiocParallelParam-class".
jobname	character(1) name (unique) used to associate manager & workers on a queue.
log	See ?"BiocParallelParam-class".
logdir	See ?"BiocParallelParam-class".
threshold	See ?"BiocParallelParam-class".
resultdir	See ?"BiocParallelParam-class".
stop.on.error	See ?"BiocParallelParam-class".
timeout	See ?"BiocParallelParam-class".
exportglobals	See ?"BiocParallelParam-class".
progressbar	See ?"BiocParallelParam-class".
RNGseed	See ?"BiocParallelParam-class".
queue.multiplier	numeric(1), The multiplier of the queue depth. The depth of the queue is calculated by queue.multiplier * bpnworkers(p). A proper queue depth can provide more performance benefit in task dispatching, but the improvement is likely to be marginal for an excessively large queue.multiplier.
redis.hostname	character(1) host name of redis server, from system environment variable REDISPARAM_HOST or REDIS_HOST, if both are not defined, the default "127.0.0.1" is used.
redis.port	integer(1) port of redis server, from system environment variable REDISPARAM_PORT or REDIS_PORT, if both are not defined, the default 6379 is used.
redis.password	character(1) or NULL, host password of redis server from system environment variable REDISPARAM_PASSWORD or REDIS_PASSWORD, if both are not defined, the default NA_character_ (no password) is used.
is.worker	logical(1) bpstart() creates worker-only (TRUE), manager-only (FALSE), or manager and worker (NA, default) connections.
x	A RedisParam object.
...	ignored.
value	The value you want to replace with

Details

Use an instance of RedisParam() for interactive parallel evaluation using bplapply() or bpiterate(). RedisParam() requires access to a redis server, running on manager.hostname (e.g., 127.0.0.1) at manager.port (e.g., 6379). The manager and workers communicate via the redis server, rather than the socket connections used by other BiocParallel back-ends.

When invoked with is.worker = NA (the default) bpstart(), bplapply() and bpiterate() start and stop redis workers on the local computer. It may be convenient to use bpstart() and bpstop() independently, to amortize the cost of worker start-up across multiple calls to bplapply() / bpiterate().

Alternatively, a manager and one or more workers can each be started in different processes across a network. The manager is started, e.g., in an interactive session, by specifying is.worker=FALSE. Workers are started, typically as background processes, with is.worker = TRUE. Both manager and workers must specify the same value for ⁠jobname =⁠, the redis key used for communication. In this scenario, workers can be added at any time, including during e.g., bplapply() evaluation on the manager. See the vignette for possible scenarios.

Value

RedisParam() returns an object of class RedisParam, for use in controlling parallel evaluation with BiocParallel::bplapply() or BiocParallel::bpiterate().

Examples

param <- RedisParam()
if (rpalive(param)) {
    res <- bplapply(1:20, function(i) Sys.getpid(), BPPARAM = param)
    table(unlist(res))
}

## Not run: 
## start workers in background proocess(es)
rscript <- R.home("bin/Rscript")
worker_script <- tempfile()
writeLines(c(
    'worker <- RedisParam::RedisParam(jobname = "demo", is.worker = TRUE)',
    'RedisParam::bpstart(worker)'
), worker_script)

for (i in seq_len(2))
    system2(rscript, worker_script, wait = FALSE)

## start manager
p <- RedisParam(jobname = "demo", is.worker = FALSE)
result <- bplapply(1:5, function(i) Sys.getpid(), BPPARAM = p)
table(unlist(result))

## stop all workers
rpstopall(p)

## End(Not run)

---