From 7b884f27060e749f017fc965c55736f72c3c719e Mon Sep 17 00:00:00 2001 From: Brooks Davis Date: Mon, 16 Mar 2009 10:17:31 +0000 Subject: [PATCH] Add an idea for groups of processes with collective resource limits (aka jobs). --- en/projects/ideas/ideas.xml | 37 ++++++++++++++++++++++++++++++++++++- 1 file changed, 36 insertions(+), 1 deletion(-) diff --git a/en/projects/ideas/ideas.xml b/en/projects/ideas/ideas.xml index 50e2e93dd3..8fbdc10439 100644 --- a/en/projects/ideas/ideas.xml +++ b/en/projects/ideas/ideas.xml @@ -15,7 +15,7 @@ Ideas//EN" - $FreeBSD: www/en/projects/ideas/ideas.xml,v 1.115 2009/03/16 00:27:38 brooks Exp $ + $FreeBSD: www/en/projects/ideas/ideas.xml,v 1.116 2009/03/16 00:38:27 brooks Exp $ @@ -712,6 +712,41 @@ href="http://info.iet.unipi.it/~luigi/FreeBSD/linux_bsd_kld.html">here. + + Collective limits on set of processes (a.k.a. jobs) + + +

Technical contact: Brooks Davis

+ +

In SGI's Irix operating system, there is a concept of a job which is + a collection of processes. These processes share a set of resource + limits similar to those accessible through the set/getrlimit and + getrusage system calls. Schedulers such as Sun Grid Engine currently + implement tracking the processes that make up a job and enforcing + collective limits on them in an adhoc manner. Having first class + kernel support would be useful.

+ +

It seems most likely that implementing something like the Irix + interface would be the most useful approach since that would enable us + to leverage existing code. Implementers will need to make sure that + the job construct has no negative performance implications when not + enabled (ideally, processes that are not part of jobs should be + unaffected) and quantify the impact on perfomance when enabled.

+ +

References: Irix + Man Pages mentioning jid_t

+ +

Requirements:

+
    +
  • Good knowledge of C.
  • +
  • Kernel awareness.
  • +
  • Ability to implement functionality from interface specifications.
  • +
+
+
+ Geom-based Disk Schedulers