1 |
Thanx for the replies, |
2 |
|
3 |
I will certainly try the beowulf mailinglist next week, |
4 |
The suggestions are great gearman seems to fit except for being perl |
5 |
oriented (atleast that's what I read sofar). |
6 |
Torque/pbs, don't know about that yet. |
7 |
|
8 |
Thanx again, |
9 |
|
10 |
Jos Houtman |
11 |
System administrator Hyves.nl |
12 |
email: jos@×××××.nl |
13 |
|
14 |
-----Original Message----- |
15 |
From: Robin H. Johnson [mailto:robbat2@g.o] |
16 |
Sent: donderdag 10 januari 2008 15:30 |
17 |
To: gentoo-cluster@l.g.o |
18 |
Subject: Re: [gentoo-cluster] cluster or distributed queue, general |
19 |
question |
20 |
|
21 |
On Thu, Jan 10, 2008 at 02:59:27PM +0100, Jos Houtman wrote: |
22 |
> For my master thesis I took up a project that requires mapping of a |
23 |
number of statically defined parallel jobs into a more dynamic |
24 |
environment that allows better scaling. |
25 |
> The situation as described below let me to believe a cluster or |
26 |
distributed queue (DrQueue?) solution is necessary. For the situation |
27 |
see [situation] at the end of this email. |
28 |
Off the top of my head, many of your requirements are available in two |
29 |
totally different apps: |
30 |
- Gearman, written by Brad Fitzpatrick @ LiveJournal. Perl mainly, I |
31 |
think there are other interfaces as well to it. |
32 |
- Torque/PBS - somewhat less of a fit, I'm not certain about running |
33 |
perpetual jobs. |
34 |
|
35 |
You may also need some degree of STONITH for the job running only once |
36 |
during node failure case. (Say the job manager crashes, the job is still |
37 |
running, but you have no control of it. You need to zap it hard). |
38 |
|
39 |
-- |
40 |
Robin Hugh Johnson |
41 |
Gentoo Linux Developer & Infra Guy |
42 |
E-Mail : robbat2@g.o |
43 |
GnuPG FP : 11AC BA4F 4778 E3F6 E4ED F38E B27B 944E 3488 4E85 |
44 |
-- |
45 |
gentoo-cluster@l.g.o mailing list |