Kolkata Tier2 @ ALICE Grid at VECC, Kolkatasymposium.vecc.gov.in/sacet09/public_html/downloads...V....

2
Kolkata Tier2 @ ALICE Grid at VECC, Kolkata V. Singhal #1 , S. Chattopadhyay #2 , T. Nayak #3 , S. K. Pal #4 , T. Samanta *5 ,Y. P. Viyogi #6 # Experimental High Energy Physics & Application Group, VECC *Computer & Informatics Group, VECC 1/AF, Bidhan Nagar, Kolkata-700064 { 1 vikas, 2 sub, 3 nayak, 4 sushant, 5 tsamanta, 6 viyogi}@veccal.ernet.in Abstract LHC will produce large amount of data, which requires huge computing power to compute & analyse the data and Peta Byte (1PB = 10 15 Bytes) order of storage to store raw and analysed data, In this regard LHC is developing WLCG (Worldwide LHC Computing Grid) which follows tier like Monarch architecture and in which Kolkata is one of the Tier-2 Centre for Alice Grid (Alice is one of the major experiment at LHC). In this connection the document will describe Alice grid in whole, its architecture, tier model, jobs of different tiers and how job submission works in Alice Grid and how storage work with xrootd. It describes Kolkata Tier2 connectivity with Alice and explains Kolkata Tier2 with its hardware, architecture and middle ware installation in whole for Cream-CE and Pure XrootD Storage. KeywordsGrid, Alice, Kolkata Tier2, CE, Pure XrootD. I. INTRODUCTION Kolkata Tier-2 has pledged to provide 3% of total ALICE Grid resources for calibrating and analysing whole raw data and then store analysed MC events. To achieve this target Kolkata Tier-2 is expanding as per the ALICE Grid requirement with respect to hardware, software and middleware. As per WLCG architecture Kolkata is a Tier-2 site for Alice Grid. II. WLCG COMPUTING MODEL WLCG grid is built on tier like architecture. Each and every site is known as one of the tier of WLCG Grid. It is categorised under 5 tiers (Tier0 to Tier4). As far as middleware is concerned tier-0,1,2 has similar installations. It differs with respect to QoS (Quality of Service) only and provides dedicated resources for the experiment as per pledged in collaboration. CERN is the Tier-0 Centre for WLCG. First & prompt calibration and reconstruction of RAW data delivered by DAQ is to be performed at Tier-0 and after that 3 kind of files are produced. Further reconstruction and calibration of RAW data is to be done at Tier-1 and Tier-2. DPD(Derived Physics Data) objects is to be processed at Tier3 & 4 sites. Tier-3 & 4 is not included as an active or live component of the Grid as far as services and resources are concerned. Tier-3, 4 are physics users with their laptop or computer on which experiment related UI is installed so they can use the Grid as per there requirement with X.509 certificate. III. JOB SUBMISSION AT ALICE GRID A user submits a job to AliEn2 (ALICE Environment is a user interface for ALICE Grid). Job is registered in JC (Job Catalogue). JC is a table which maps jobs with its respective LFNs (logical file names). Then respective GUIDs (Grid Unique Identifiers) are mapped with LFN as per File Catalogue and respective SEs (storage element) are mapped with the LFNs. The whole process is done by Central Services. According to closest SE and software matchmaking, job is submitted to respective Grid site. After this job agent is submitted to WMS. Fig. 1 Alice Job Submission procedure described through AliEn Central and site services. WMS sends job agent to site CE. Job scheduler running on CE submits the job to available WN (Worker Node). If Environment of the WN is not proper then job agent dies and reports to scheduler. If Environment OK then JA start on WN and install required packages after consultation with packman service running on VOBOX and asks all related files to CE service. All related files are fetched from respective SE as per JC. After completion job results sends to Alien-CE. Now accordingly Task Queue is updated and output is registered into ALICE FC (File Catalogue). IV. KOLKATA TIER-2 DEVELOPMENT Kolkata Tier-2 has 2 clusters, one named KOLKATA (in Alien ldap) is configured with LCG-CE (Computing Element) on 32 bit SLC4.6 (i386) comprises 13 Nos of Dual Xeon 3.0GHz WIPRO Netpower servers equivalent of 30 KSI2k (SpecInt2000) and another one named Kolkata-Cream is configured with Cream-CE (Computing Resource Execution

Transcript of Kolkata Tier2 @ ALICE Grid at VECC, Kolkatasymposium.vecc.gov.in/sacet09/public_html/downloads...V....

Page 1: Kolkata Tier2 @ ALICE Grid at VECC, Kolkatasymposium.vecc.gov.in/sacet09/public_html/downloads...V. KOLKATA SE BASED ON PURE XROOTD Pure XrootD is a protocol which enables file access,

Kolkata Tier2 @ ALICE Grid at VECC, Kolkata V. Singhal#1, S. Chattopadhyay#2, T. Nayak#3, S. K. Pal#4, T. Samanta*5,Y. P. Viyogi#6

#Experimental High Energy Physics & Application Group, VECC *Computer & Informatics Group, VECC

1/AF, Bidhan Nagar, Kolkata-700064 {1vikas, 2sub, 3nayak, 4sushant, 5tsamanta, 6viyogi}@veccal.ernet.in

Abstract — LHC will produce large amount of data, which requires huge computing power to compute & analyse the data and Peta Byte (1PB = 1015 Bytes) order of storage to store raw and analysed data, In this regard LHC is developing WLCG (Worldwide LHC Computing Grid) which follows tier like Monarch architecture and in which Kolkata is one of the Tier-2 Centre for Alice Grid (Alice is one of the major experiment at LHC). In this connection the document will describe Alice grid in whole, its architecture, tier model, jobs of different tiers and how job submission works in Alice Grid and how storage work with xrootd. It describes Kolkata Tier2 connectivity with Alice and explains Kolkata Tier2 with its hardware, architecture and middle ware installation in whole for Cream-CE and Pure XrootD Storage. Keywords— Grid, Alice, Kolkata Tier2, CE, Pure XrootD.

I. INTRODUCTION Kolkata Tier-2 has pledged to provide 3% of total ALICE Grid resources for calibrating and analysing whole raw data and then store analysed MC events. To achieve this target Kolkata Tier-2 is expanding as per the ALICE Grid requirement with respect to hardware, software and middleware. As per WLCG architecture Kolkata is a Tier-2 site for Alice Grid.

II. WLCG COMPUTING MODEL WLCG grid is built on tier like architecture. Each and every site is known as one of the tier of WLCG Grid. It is categorised under 5 tiers (Tier0 to Tier4). As far as middleware is concerned tier-0,1,2 has similar installations. It differs with respect to QoS (Quality of Service) only and provides dedicated resources for the experiment as per pledged in collaboration. CERN is the Tier-0 Centre for WLCG. First & prompt calibration and reconstruction of RAW data delivered by DAQ is to be performed at Tier-0 and after that 3 kind of files are produced. Further reconstruction and calibration of RAW data is to be done at Tier-1 and Tier-2. DPD(Derived Physics Data) objects is to be processed at Tier3 & 4 sites. Tier-3 & 4 is not included as an active or live component of the Grid as far as services and resources are concerned. Tier-3, 4 are physics users with their laptop or computer on which experiment related UI is installed so they can use the Grid as per there requirement with X.509 certificate.

III. JOB SUBMISSION AT ALICE GRID A user submits a job to AliEn2 (ALICE Environment is a user interface for ALICE Grid). Job is registered in JC (Job Catalogue). JC is a table which maps jobs with its respective LFNs (logical file names). Then respective GUIDs (Grid Unique Identifiers) are mapped with LFN as per File Catalogue and respective SEs (storage element) are mapped with the LFNs. The whole process is done by Central Services. According to closest SE and software matchmaking, job is submitted to respective Grid site. After this job agent is submitted to WMS.

Fig. 1 Alice Job Submission procedure described through AliEn Central and site services. WMS sends job agent to site CE. Job scheduler running on CE submits the job to available WN (Worker Node). If Environment of the WN is not proper then job agent dies and reports to scheduler. If Environment OK then JA start on WN and install required packages after consultation with packman service running on VOBOX and asks all related files to CE service. All related files are fetched from respective SE as per JC. After completion job results sends to Alien-CE. Now accordingly Task Queue is updated and output is registered into ALICE FC (File Catalogue).

IV. KOLKATA TIER-2 DEVELOPMENT Kolkata Tier-2 has 2 clusters, one named KOLKATA (in Alien ldap) is configured with LCG-CE (Computing Element) on 32 bit SLC4.6 (i386) comprises 13 Nos of Dual Xeon 3.0GHz WIPRO Netpower servers equivalent of 30 KSI2k (SpecInt2000) and another one named Kolkata-Cream is configured with Cream-CE (Computing Resource Execution

Page 2: Kolkata Tier2 @ ALICE Grid at VECC, Kolkatasymposium.vecc.gov.in/sacet09/public_html/downloads...V. KOLKATA SE BASED ON PURE XROOTD Pure XrootD is a protocol which enables file access,

and Management) on 64bit SLC4.6 (x86_64) comprises 39 Nos of 2*Dual Core Xeon 3.0 GHz HP Blade servers equivalent of 250 KSI2k. Both sides are registered as IN-DAE-VECC-01 and IN-DAE-VECC-02 in GOCDB (Grid Operation Centre Database).

Fig. 2 Full logical diagram of Kolkata Tier2 Site which comprises of 3 clusters. (The 5 node grid-peer cluster is used as tier-3 for local users)

In Grid Architecture network is a critical component (as computing jobs run globally and then result stores on the SE which is distributed across the globe.) Kolkata Centre has 100Mbps dedicated network speed which is used for fast file transfer across the all ALICE Grid sites. 3 * 40 KVA UPS power supply is arranged for providing uniform power to the centre and cooling infrastructure is also arranged as per the site resources requirement with the help of VECC, Electrical and AC Division.

Fig. 3 MonaLisa Repository Map shows both Kolkata clusters are up and running with its maximum present capacity and data transfer with different Alice sites are going on.

V. KOLKATA SE BASED ON PURE XROOTD Pure XrootD is a protocol which enables file access, replication and uploading of file on geographically distributed storage element irrespective of type of storage. It has 2 types of servers one Redirector (head node) which maps files to its disk servers other one is Disk/Data Servers which physically consists the files. Presently Kolkata SE has one redirector named dcache-server and 2 disk servers names dcache-pool, dcache-client consisting of HP 2 * EVA6100 with 108TB

Raw disks (75TB usable space after Global hot spare and RAID5 configuration).

Fig. 4 Pure XrootD SE architecture shows how a file is retrieved via xrootd.

Fig. 5 HP 2 * EVA 6100 with 108TB Raw Disks

VI CONCLUSIONS Kolkata Tier-2 successfully added 250 K SI2k (SpecInt2000) computing power and 75TB usable storage. Kolkata Tier-2 is continuously participating in all Physics Data Challenges (like PDC04, PDC06, PDC07, PDC08 etc.) and also successfully completed CCRC08 (Common Computing Readiness Challenge before starting LHC). As per the EGEE Availability and Reliability report (based on SAM and GridView database), reliability and availability factor for Kolkata Tier-2 is more than 95%.

ACKNOWLEDGMENT Kolkata Tier-2 comprises good amount of servers, storage and 100Mbps network bandwidth, the all procurement is supported from the grants of the Department of Atomic Energy and Department of Science and Technology, Government of India. I would like to thank the entire members of ALICE Grid, CERN and APROC Taiwan who helped in deployment of Grid middleware at Kolkata site.

REFERENCES

[1]. P. M. Lorenzo, IT/GS, CERN, “VOBOXes and LCG Services” in Alice T1/T2 Tutorial for Site Admins, CERN, Geneva 26-27May 2009. [2]. L. Betev, Alice, CERN, “Grid Services and relations” in Alice T1/T2 Tutorial for Site Admins, CERN, Geneva 26-27May 2009.