1. INRNE's participation in LCG Elena Puncheva (elly@inrne.bas.bg) Preslav Konstantinov (pkonst@inrne.bas.bg) IT Department

2. Motivation ✔ To make available adequate computing facilities such that the Bulgarian HEP scientists can contribute fully to the physics programme of LHC and to other experiments. ✔ The scientific community at large would also benefit from the available resources.

3. First steps ✔ Cooperation agreement with CERN on "Development of Compute Fabrics for High-Energy Physics and other Sciences", 2001. ✔ CERN donates machines phased out of its fabrics and provides training ✔ INRNE provides the necessary infrastructure and equipment ✔ A local facility is put in operation at INRNE BAS, 2002. Available for use by staff from other institutes, too.

4. First steps (cont.) ✔ The compute fabric consists of 12 dual processor WNs, 1 interactive node and 2 server nodes ✔ Experience is gained in deployment and management of a cluster ➢ automatic, script-driven installation and configuration ➢ using LDAP to store common accounts info and automount maps ➢ PBS and Maui used for resource management

5. The present... ✔ A second cluster, running LCG2 middleware is deployed and maintained. ✔ LHC Computing Grid site since the beginning of the year. ✔ Consists of CE, SE, MON and 10 WNs. All dual processor. ✔ Number of WNs to be doubled and a 1 TB disk server to be added by the end of the year.

6. Features ✔ Network design with WNs on a private network. ➢ benefits are conservation of public IP address space and no packet filtering on the WNs ➢ outbound connectivity for WNs provided by a separate NAT box ➢ Grid service nodes connected to both public and private networks ➢ network segments implemented as 802.1q VLANs using support in Linux kernel ➢ traffic local to the site (job submission, monitoring, etc.) is confined to the private network

7. NATed WNs WN CESEMONNAT Private VLAN Public VLAN

8. Features ✔ Network design with WNs on a private network. ➢ benefits are conservation of public IP address space and no packet filtering on the WNs ➢ outbound connectivity for WNs provided by a separate NAT box ➢ Grid service nodes connected to both public and private networks ➢ network segments implemented as 802.1q VLANs using support in Linux kernel ➢ traffic local to the site (job submission, monitoring, etc.) is confined to the private network ✔ Direct access to the LRMS for local job submission ➢ 1.4 TB disk server for CMS software and data storage ➢ scientists from CMS group use their own accounts ➢ separate queue is used for local jobs

9. Training ✔ Visits by INRNE staff to CERN IT to become familiar with the techniques used for deployment of compute fabrics. ✔ Involvement in projects of the Fabric and Infrastructure Operations group ✔ developed a prototype web application for displaying data collected by LEMON monitoring agents in the Measurement Repository ✔ performed a Quattor installation from scratch and provided feedback to the developers in order to improve the installation manual

10. Thank you!