Next: Front End Utilities
Up: 1993 EXPERT'S GUIDE ONLINE
Previous: THE MESSAGING SYSTEM
The front end code uses socket-driven inter-process communications this
year, in order to accomodate the new HP Filter machine, and to facilitate
hardware migration in online ROPE. There are two main server-client models
used on the front-end:
- This year the Filter is an HP 742/747 super workstation. The IP name of
this node (on public ethernet) is hpop7f2. You can telnet into this
node by specifying this IP name, and by logging in as user filter. Do
not maintain unused connections on this machine, and do not run anything
that is not strictly needed.
- Dense data files from either online or offline
processes are distributed from the RopeServer to any RopeClient process
that requests a file. As described in the next section, the RopeClient always
seeks an online file first. This assures that the online processing will
get first priority.
The RopeServer maintains a state table of the
ROPE engines that it knows about. This information can be used to
understand the general state of ROPE processing, and the health of
a ROPE engine. RopeClients communicate three kinds of information
to the server - a request for a data file, a change of state or
a notification that it has finished a file. The states that the
server understands are:
Note that in test states the ROPE engine only wants test files.
- ONLINE or ONLINE_TEST
- OFFLINE or OFFLINE_TEST
- IDLE - the node is shutting ROPE off
- FAULT - the node has a disk space problem
The server also has knowledge about the past 10 files that a ROPE
engine has dealt with.
Note about online verses offline running.