Re: [HTCondor-users] Designing a new HTC cluster

[Christopher Harrison <harrison@xxxxxxxxxxxxxxxx>]

Good question.   I have been expanding our HTC pool on the cheap for a while.   I would be happy to discuss further with you over chat.   Let me know some times you have available to talk.

-C

---

From: HTCondor-users <htcondor-users-bounces@xxxxxxxxxxx> on behalf of mtwest2718 via HTCondor-users <htcondor-users@xxxxxxxxxxx>
Sent: Thursday, October 8, 2020 5:34 PM
To: htcondor-users@xxxxxxxxxxx <htcondor-users@xxxxxxxxxxx>
Cc: mtwest2718 <mtwest2718@xxxxxxxxxxxxxx>
Subject: [HTCondor-users] Designing a new HTC cluster

 

While I realize a bunch of this would be worked out with the hardware vendor, do any of the sysadmins on here have recommendations on how to design a new system given a set of monetary and spatial constraints? I figure there has to be some sort of "configurator" out there.

The reason I ask is that I would like to advocate for a potential new system to be built for HTC workloads and would need some concrete throughput per dollar numbers to justify this change over just expanding the existing HPC system. And who better to ask than the community that does so much to advocate for high-throughput methods internationally?

Cheers,
Matt