Quantcast
Channel: Embedded Community : All Content - All Communities
Viewing all articles
Browse latest Browse all 3140

Dynamic Hugepage allocation for DPDK

$
0
0

On the DPDK.org website it doesn't provide a very good explanation why 2048 MB of hugepages is allocated or when that size should be increased or could be decreased.  I am trying to understand why the requirements are what they are and if the explanation is either oversimplified or simply is that way for a reason.

 

I will list my questions in order.

 

1. NUMA - If all DPDK compatible interfaces are on a single NUMA node then the hugepages should only be allocated for that NUMA node?

2. NUMA - If interfaces are scattered on node 0 and 1 are the basic reccomendations to allocate 1024 hugepages per node or 512 per node?

3. NICS - Does the hugepage size allocation change if we are using the DPDK only with a single 1 Gb interface or 4x 10 Gb interfaces?

 

I think this is a good start and possibly there might be a few follow up questions.


Viewing all articles
Browse latest Browse all 3140

Latest Images

Trending Articles



Latest Images

<script src="https://jsc.adskeeper.com/r/s/rssing.com.1596347.js" async> </script>