You are viewing a plain text version of this content. The canonical link for it is here.
Posted to yarn-issues@hadoop.apache.org by "Adam Antal (Jira)" <ji...@apache.org> on 2021/02/12 14:14:00 UTC

[jira] [Commented] (YARN-10621) GPU management using OpenCL instead of vendor-specific solutions

    [ https://issues.apache.org/jira/browse/YARN-10621?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17283717#comment-17283717 ] 

Adam Antal commented on YARN-10621:
-----------------------------------

Thanks for bringing this issue to the community!

From the JIRA details I assume that you're using version 3.1.1, which has not yet included the {{{DevicePlugin}} interface. This interface has been added to the code for the very same purpose: discovering custom resources provided by these plugin - just like the Nvidia GPUs. For more information look at the umbrella jira: YARN-8851.

I don't know if the changes you've been working on are based on this work, but it's the recommended way from 3.3.0 on. How is that aligns with you effort?


> GPU management using OpenCL instead of vendor-specific solutions
> ----------------------------------------------------------------
>
>                 Key: YARN-10621
>                 URL: https://issues.apache.org/jira/browse/YARN-10621
>             Project: Hadoop YARN
>          Issue Type: Improvement
>          Components: nodemanager, yarn
>            Reporter: Sotiris Niarchos
>            Priority: Minor
>
> As part of the [E2Data research project|https://e2data.eu/], we at the [Institute of Communication and Computer Systems (ICCS)|https://www.iccs.gr/en/?noredirect=en_US] of the National Technical University of Athens, Greece, have been working on a modified version of Hadoop Yarn where the GPU devices that are available in the underlying cluster are discovered via a Java wrapper of the OpenCL framework API (namely [JOCL|https://github.com/gpu/JOCL]), instead of vendor-specific binaries.
> In other words, we have shifted towards *a more uniform and high-level handling of GPUs as "OpenCL-enabled" devices*. This way, we manage to *decouple GPU discovery/management from vendor-specific technicalities*; every GPU, no matter the vendor, is the same for E2Data YARN (more specifically, for the {{NodeManager}} component), provided that the OpenCL runtime and drivers for the GPU(s) of interest are installed on the respective node(s) of the cluster.
> This way, we *managed to use GPUs other than NVIDIA* (which are the only ones officially supported via the {{nvidia-smi}} binary) with minimal additional effort, after our initial changes.
> Ultimately, our goal is to *unify every processing unit* that YARN can possible utilize (CPU cores, GPUs, FPGAs) *behind a common, simple, high-level interface; that of the OpenCL-enabled device*.
> The only drawback of our approach is that vendor-specific info regarding the GPUs is lost (e.g. temperature). We believe, however, that the lost information is not necessary for YARN; everything that Hadoop needs in order to discover and handle GPU devices is provided by OpenCL.
> This is just a proposition/a prompt for discussion for the time being. This modified version is a work in progress. We consider community feedback regarding the core concept (and the fact that it may constitute a paradigm shift for YARN) crucial before attaching any patch file and diving into more (technical) details.



--
This message was sent by Atlassian Jira
(v8.3.4#803005)

---------------------------------------------------------------------
To unsubscribe, e-mail: yarn-issues-unsubscribe@hadoop.apache.org
For additional commands, e-mail: yarn-issues-help@hadoop.apache.org