Technical note Open Access
{ "DOI": "10.15161/oar.it/143357", "abstract": "<p>The default behaviour of a batch system is to dispatch jobs to nodes having the lower<br>\nvalue of some load index. Whilst this causes jobs to be equally distributed among all the<br>\nnodes in the farm, there are cases when different types of behaviour may be desirable, such as<br>\nhaving a completely full node before dispatching jobs to another one, or having similar jobs<br>\ndispatched to nodes already running jobs of the same kind. This work defines the packing<br>\nconcept, different packing policies and useful metrics to evaluate how good the policy is. A<br>\nsimple farm simulator has been written to evaluate the expected impact on a farm of different<br>\npacking policy. The simulator is run against a sequence of real jobs, whose parameters have<br>\nbeen taken from the accounting database of INFN-Tier1. The effectiveness of two packing<br>\npolicies of interest, namely relaxed and exclusive, are compared. The exclusive policy proves<br>\nto be better, at the cost of unused cores in the farm, whose number is estimated. The<br>\npossibility of implementing the exclusive policy on a specific batch system, LSF 7.06, is<br>\nexploited. Relevant configurations are shown and an overall description of the mechanism is<br>\npresented.</p>", "author": [ { "family": "Dal Pra, Stefano" } ], "id": "143357", "issued": { "date-parts": [ [ 2014, 12, 16 ] ] }, "language": "eng", "title": "JOB PACKING: OPTIMIZED CONFIGURATION FOR JOB SCHEDULING", "type": "article" }
All versions | This version | |
---|---|---|
Views | 0 | 0 |
Downloads | 0 | 0 |
Data volume | 0 Bytes | 0 Bytes |
Unique views | 0 | 0 |
Unique downloads | 0 | 0 |