GXP Grid & Cluster Shell

NEWS


GXP is a parallel shell tool that can work on any machine reachable via one of these "remote-exec" commands (ssh, rsh, batch queueing commands, etc.). Machines may be in your office LAN, in a cluster, or in multiple LANs/clusters widely distributed. These machines may be behind NAT routers or firewalls, as long as you can reach them via multiple hops of remote-exec commands (e.g., ssh to a gateway and then to hosts behind firewalls). No matter where they are, GXP connects all reachable machines with underlying remote-exec commands and lets you operate them very efficiently. With a single stroke of typing a command, you can invoke it on hundreds of computers in seconds (demo video).

What you can do with GXP

Basically you can run an identical or a similar command line to many machines in parallel and get results back interactively. You can choose host to execute commands on interactively by various criterion. Tools accompanying GXP allow you to run embarrassingly parallel jobs on any machine reachable via GXP.

Some notable features

Easy to install
When you use GXP on many hosts (as you probably do), you need to install it only on one host. GXP automatically copies itself to remote hosts you run commands on, as necessary. It is written in Python so installation only takes download + PATH setting (plus whatever your underlying remote-exec commands demand for successful authentication).
Fast in distributed environments
GXP brings up daemon processes on hosts you run commands on and keeps them running until you quit a session. It does not require connection establishment and authentication for every single command submission. This feature makes operations fast especially in distributed environments where some hosts are remote, and in large scale environments having many hosts.
Scalable
Daemon processes are connected in a tree structure so as to keep the number of processes directly connected to a single process below a constant threshold. Thus GXP comfortably scales to hundreds of processes without stressing the root host.
Interactive
GXP allows you to interactively specify hosts on which you bring up daemons and then select hosts to run commands on. It keeps these configurations in a session state. In a simple cluster-like enviroment, you typically need only a few interactions with GXP to run commands in parallel on cluster hosts. In more complex and heterogeneous environments, you can write whatever configurations are necessary just as a shell script you are familar with.
Useful for parallel processing
GXP accompanies a tool to perform so-called "embarrassingly parallel tasks" with a trivial amount of learning and setting. It does not need any other infrastructure (e.g., batch queueing system).

today 2 / total 45525 accesses


トップ   編集 バックアップ 添付 複製 名前変更 リロード   新規 一覧 最終更新   最終更新のRSS
© 2007 Taura lab. All Rights Reserved.
Powered by Pukiwiki, styled by Kei_ / Last-modified: 2016-07-12 (火) 15:12:44 (48d)