 |
 |
Hi On Wed, Oct 7, 2009 at 9:37 AM, Alessandra Forti <[log in to unmask]> wrote: > From bnl site (follow the links) these are the raccomandations for hardware > > These recommendations are based on Frontier Squid for CMS, but they have > proven adequate thus far in US ATLAS testing. > > * 200+ GB disk (for cache and logging) > * 2 GB memory > * Gigabit Ethernet interface > * Dual core CPU > /Squid is single-threaded, but it can consume 100% of a CPU core > under heavy load/ > * 64-bit OS > Unless the ATLAS usage pattern is wildly different from the cms one, we use nowhere near these resources. The imperial one is used by all the uk cms tier2/3's apart from Oxford and RAL and see's very little load. Its a vm with specs of: cpu_num 2 CPUs cpu_speed 1994 MHz mem_total 1048576 KB swap_total 2097144 KB the only resource use we really see is bandwidth of upto ~10MB/s and disk usage from logs which can fill GB's a day. CPU has never gone above ~5% from a quick look at ganglia. Would be interesting to know what circumstances the 100% cpu was caused by. RAL's cms squids: http://ganglia.gridpp.rl.ac.uk/ganglia/?m=load_one&r=hour&s=descending&c=Services_Grid&h=lcgvo0428.gridpp.rl.ac.uk&sh=1&hc=4 http://ganglia.gridpp.rl.ac.uk/ganglia/?m=load_one&r=hour&s=descending&c=Services_Grid&h=lcgvo0599.gridpp.rl.ac.uk&sh=1&hc=4 IC's ganglia isn't visible from outside college. Cheers Stuart > > > Christopher J. Walker wrote: >> >> Graeme Stewart wrote: >>> >>> If you are an ATLAS T2 you should have already heard about this, but >>> if you haven't... >>> >>> Peter and I can advise on any technical points in the UK. >>> >> >> Will an old 32 bit machine with 2Gig ram be up to the task? >> >> Chris >> >> >>> CHeers >>> >>> Graeme >>> >>> >>> ---------- Forwarded message ---------- >>> From: Dario Barberis <[log in to unmask]> >>> Date: Tue, Oct 6, 2009 at 20:47 >>> Subject: Deployment of Frontier/Squid for ATLAS >>> To: ADC Developments <[log in to unmask]>, >>> [log in to unmask], "atlas-database (Atlas - Database >>> Group)" <[log in to unmask]>, "atlas-tier1-coordination >>> (Coordination of ATLAS Tier-1 sites)" >>> <[log in to unmask]>, "atlas-sw-tier2-contacts (ATLAS >>> Tier-2 and Analyse Facility contacts)" >>> <[log in to unmask]>, Flavia Donno >>> <[log in to unmask]>, Maria Girone <[log in to unmask]> >>> Cc: ATLAS CMB <[log in to unmask]>, [log in to unmask] >>> >>> >>> Dear all, >>> >>> ATLAS has decided to move ahead with a rapid deployment of FroNTier >>> servers and Squid web caches to address the shortcomings of direct >>> remote access to Oracle databases holding conditions data that are >>> needed by jobs running worldwide. See note ATL-SOFT-INT-2009-003 >>> (http://cdsweb.cern.ch/record/1210179/files/ATL-SOFT-INT-2009-003.pdf) >>> for more details. >>> The report of the "Analysis Model for the First Year" (AMFY) Task >>> Force presented at today's ATLAS plenary meeting supports this rapid >>> deployment; see >>> >>> http://indico.cern.ch/getFile.py/access?contribId=13&sessionId=6&resId=1&materialId=slides&confId=47256 >>> for more details. >>> >>> Thanks to the development and deployment work of the BNL and other >>> US-ATLAS sites, and the testing done in the German cloud and at CERN >>> with help also from CMS experts, we now have the tools for a quick >>> general deployment. >>> It is required that every Tier-1 and Tier-2 site install a Squid server >>> asap. >>> A number of Tier-1s have installed, or are going to install, a >>> FroNTier server; we are having, or going to have soon, direct >>> discussions with Tier-1 sites. A FroNTier server is also under test at >>> CERN, but this server will be initially reserved for further load >>> tests. >>> Preliminary instructions can be found in >>> https://twiki.cern.ch/twiki/bin/view/Atlas/T2SquidDeployment and the >>> references therein. While more FroNTier servers are installed, the >>> functionality of Squids can be tested immediately against the BNL >>> and/or FZK servers. The same wiki page is used to keep track of >>> installation and test progress. >>> >>> In order to speed up the deployment of the Frontier/Squid system for >>> ATLAS, we appoint John DeStefano (BNL) and Rod Walker (LMU) as >>> coordinators of this activity. Within this organization, Douglas Smith >>> (SLAC) will be the reference person concerning Squid issues. >>> For general questions and hopefully answers, sites should send >>> messages to the database operations support list at >>> [log in to unmask] >>> Dario Barberis & Kors Bos & Jim Shank >>> ps: apologies for this multiple posting. >>> ______________________________________________________________ >>> Dr Dario Barberis >>> CERN-PH Department Tel.: +41.22.767.1302 >>> MailBox E25510 Fax.: +41.22.767.8350 >>> CH-1211 Genève 23 (Switzerland) >>> ______________________________________________________________ >>> >>> >>> >>> >>> > > -- > Mindmelds. The last time I heard the words "my mind to your mind", I had a > headache for two weeks. (Janeway, ST Voyager) > > Northgrid Tier2 Technical Coordinator > http://www.hep.manchester.ac.uk/computing/tier2 > >