1 INDIACMS-TIFR TIER-2 Grid Status Report IndiaCMS Meeting, Sep 27-28, 2007 Delhi University, India.
-
Upload
blake-floyd -
Category
Documents
-
view
216 -
download
0
Transcript of 1 INDIACMS-TIFR TIER-2 Grid Status Report IndiaCMS Meeting, Sep 27-28, 2007 Delhi University, India.
1
INDIACMS-TIFR TIER-2 Grid Status Report
IndiaCMS Meeting, Sep 27-28, 2007
Delhi University, India
2
Site Information Site name: INDIACMS-TIFR Site address:
http://www.indiacms.res.in Email: [email protected] User Interface: ui.indiacms.res.in Job submission Queues:
ce.indiacms.res.in:2119/jobmanager-lcgpbs-dteam ce.indiacms.res.in:2119/jobmanager-lcgpbs-cms
3
Current Status -CPUs CPU Integer Processing Power
80k SPECInt2000 equivalent 1U Form Factor Rack mounted Dual Xeon/2GB/80GB(Total 24 servers)
1 Storage Element 1 Computing Element 22 Worker nodes
Connected to GigaBit and KVM switches Latest rpms installed on managing servers and
worker nodes
4
Current Status - Disk Disk Storage
50 TB 33x300 GB Fast Storage 80x500 GB Slow Storage
Storage Array HP EVA 8000
Fully populated single rack mount DPM (SRM support) installed Gateway Server configured No propritory software
5
Current Status - Network Connectivity
Bandwidth Current bandwidth of 34MBps will be increased to
1 GBps eventually. Switches –2
48-port Gigabit switch for servers to connect to internet and between themselves.
48-port Gigabit switch for private LAN. GBIC Transreceiver
Gigabit interface converter(DEM-310 GT) for the Gigabit switch used with servers.
Point to Point connection (1:1)
6
Software Updates Middleware
gLite 3.0.2 + Latest packages installed
Operating System + Software's Worker Nodes
Support for SL 3.x will be withdrawn Upgraded to SL 4.4
Server Nodes Still on SL 3.x Support for SL 4.x is not yet available.
7
Infrastructure Existing UPS
20KVA UPS with 3 Hrs Battery Backup Batteries are rack mounted
Additional capacity Work order is ready to bring additional 40KVA capacity to
the room Indent for second UPS is ready – 40KVA, 3hr batter backup
full load. Central AC upgrade
has been finalized with Central Services of TIFR. Floor plan for full capacity
Ready - indicates power outlets, rack positions etc.
8
Site resources as in MOU Additional resources are now being procured to bring the Site up to the
promised sizes. Storage
Additional 400TB of raw disk space Indent is ready – aiming for smaller size enclosures, with 1TB SATA
HDDs, and multiple controllers. Aiming to reduce floor space requirement, power consumption, and AC
load Tender Being Prepared.
CPU Integer Processing Power Additional required - 360K SPECInt2000 Considering Quad Core Dual CPU Blade Servers Public Tender Floated for 45 Blade servers
UPS Additional 40 KVA UPS for increased Power Tender Being Floated.
9
PhEDEx Installation and Testing PhEDEx (Physics Experiment Data Export) - it is a Data
Transfer and Placement system designed for CMS Used to manage data flow, following a transfer topology of
Tier0 -> Tier1 -> Tier2 Is installed, typically, on an UI on same domain as the SE Operated by a person(s) designated as “CMS Data
Manager” Is a set of daemons to manage reliable data xfers, removal,
export, etc., both in Debug mode and Prod mode. Another key component – the Trivial File Catalog (set of
rules to decide which datasets go into which dir in the SE)
10
PhEDEx at INDIACMS Installed on ui.indiacms.res.in Phedex node name
T2_INDIACMS_TIFR Monitoring site
http://cmsdoc.cern.ch/cms/aprom/phedex/debug/Components::Status?view/==
TFC is written, currently to seperate LoadTest data from regular CMS data – will be fine tuned as understanding improves.
Download agents are running. Currently only Debug instance is being used for testing/debugging/loadtest
Subsequently the “Prod” instances will start their work.
11
Tests with basic file xfer Basic file transfers are working – with xfers
with srmcp (i.e, not with phedex agents) Tested and found ok by ROC Taiwan Taiwan reported timeout problems -
increasing timeout peroid, and number of parallel xfers did not help.
xfers from CERN OK, says Gavin (1GB files).
12
Status of PhEDEx installation All agents up and running (even now when site is under
shutdown) At monitoring site all PhEDEx components are shown
green=OK. Could subscribe to LoadTest data prepared for us(two
samples – from ASGC and from CERN) But Transfer does not happen ( PhEDEx assumes there
is problem). No other problem seen, says expert Chia-Ming Kuo.
maybe bandwidth choke up, again.
13
What now? Phedex download “may work, please try after increasing
priority of subscribed data from Normal to High”, suggest experts. (maybe allowed for us only for test/debug)
Meanwhile, shutdown because of AC, till today (27th)., and for possible network upgrade.
Phedex testing will be resumed today (27th) after lifing shutdown.
Despirately waiting for the 1GbPS link - only big change to improve things – maybe 10 days away (according to information obtained evening of 26/Sept)