Quick sizing guide for SAS Grid running on HP BladeSystems and EVA Storage

Q uick sizing guide for SAS Grid
running on HP Bla deSystems a nd
EVA Stora ge
Executive summa ry ............................................................................................................................... 2 Solution definition ................................................................................................................................ 2 Reference Configura tion ....................................................................................................................... 3 Sizing considera tions a nd a ssumptions ................................................................................................... 5 Sa mple configura tions .......................................................................................................................... 5 For more informa tion ............................................................................................................................ 8 Executive summa ry
Toda y most customers using PC SAS ha ve no control over where the da ta resides. It ma y be on PC ha rd drives, externa lly a tta ched drives (e.g. USB), persona l LAN drives or a nywhere else users ca n store da ta . O ne of the drivers for moving from a PC environment to a centra lized server environment is a n a ttempt to get control of the da ta a nd where it resides. A solution to this issue is SAS Grid Ma na ger which a llows the centra liza tion a nd sha ring of da ta . SAS Grid Computing is a n enterprise-cla ss solution tha t ena bles SAS a pplica tions to ta ke better a dva nta ge of computing resources. SAS Grid Ma na ger ma na ges the SAS workloa d dyna mica lly a cross a ll nodes in the SAS Computing Grid. It provides dyna mic loa d ba la ncing, job prioritiza tion a nd SAS job control. It is the intent of this PC Consolida tion Q uick Sizing Guide to sha re a ssumptions a nd experiences a s the number of users is increa sed. Target audience: The ta rget a udience for this sizing guide is the IT community eva lua ting a migra tion to a SAS Grid environment. Business users a nd IT professiona ls who a re interested in implementing SAS Grid Ma na ger ma y find this sizing guide useful a s a reference rela ting to the type a nd qua ntity of equipment needed to deploy this solution for the specified number of end users. This white pa per describes testing performed in O ctober 2007.
Solution definition
SAS Grid Computing ena bles the implementa tion of SAS to proceed in a sea mless ma nner. As users a re a dded to the implementa tion it is a n ea sy ta sk to a dd a bla de a nd sca le the a pplica tion. If a user ha s a job tha t is urgent, it ma y be possible to a llow tha t job to be divided a nd run a cross more tha n one processor, something tha t running with ma ny sepa ra te PCs would ma ke difficult, if not impossible. Reference Configura tion
Figure 1. Deployment dia gra m

This reference a rchitecture this Q uick Sizing Guide is ba sed on is ba sed on testing with rea l-world da ta volumes a nd structures a s well a s experience with rea l customer SAS Grid Computing implementa tions. The SAS-developed workloa d simula ted the types of jobs received from va rious SAS clients such a s displa y ma na ger, ba tch, SAS® Da ta Integra tion Studio, SAS® Enterprise MinerT, SAS Microsoft® Add-In a nd SAS® Enterprise Guide®. Ma ny customer environments ha ve la rge numbers of a d-hoc SAS users or jobs tha t utilize a na lytics in support of their compa ny's da y to da y business a ctivities. The test scena rio represented simulta neous users tha t replica ted a mix of light to hea vy users performing typica l a na lytic jobs designed to replica te a hea vy workloa d. The SAS® Grid Ma na ger workloa d used for this test ha d the following cha ra cteristics: . 50% CPU intensive jobs and 50% I/O intensive jobs . Utilized SAS procedures including: - SAS DATA step, PRO C RISK, PRO C LO GISTIC, PRO C GLM (genera l linea r model), PRO C REQ , PRO C SQ L, PRO C MEANS, PRO C SUMMARY, PRO C FREQ a nd PRO C SO RT . SAS program input sizes up to 50 GB per job* . Input data types: text, SAS dataset and SAS tra nsport files . Memory use per job up to 1 GB . Each SAS job utilized approximately 25 MB/s I/O throughput . Job runtimes were varied (short and long running tasks)
* File sizes were chosen to eliminate the caching eff... [download for more]