Update on ComputingCloud Marco Destefanis Universit degli Studi

Update on Computing/Cloud Marco Destefanis Università degli Studi di Torino Stefano Bagnasco, Flavio Astorino, Dario Berzano, Stefano Lusso, Marco Maggiora, Sara Vallero, Laura Zotti IHEP Beijing BESIII Ferrara, Italy October 21, 2014 1

Motivation • The amount of resources and the variety of applications is steadily increasing, manpower unfortunately is not • It is becoming almost mandatory to consolidate such resources to achieve scalability and economies-of-scale – Separate application management from infrastructure management – Our Data Centers need to become providers of computing and storage resources, not (only) of high level services • The Cloud approach (Iaa. S) might help to better provision resources to the different scientific computing applications – Grid Sites, small or medium computing farms, single users, … – Admit dynamic resource relocation to increase CPU power for a Grid and reduce some other that are not using resources or having less priority • Several cloud computing projects are starting at national and European level – From definition of best practices and reference configurations to deployment of large-scale distributed infrastructures – A local working cloud infrastructure will also allow to take immediately part in such activities

Cloud Computing • On-demand self-service. – A consumer can unilaterally provision computing capabilities, such as server time and network storage, as needed automatically without requiring human interaction with each service provider. • Broad network access. – Capabilities are available over the network and accessed through standard mechanisms that promote use by heterogeneous thin or thick client. • Resource pooling. – Computing resources are pooled to serve multiple consumers using a multi-tenant model, with different physical and virtual resources dynamically assigned and reassigned according to consumer demand. • Rapid elasticity. – Capabilities can be elastically provisioned and released, in some cases automatically, to scale rapidly outward and inward commensurate with demand. • Measured service. – Cloud systems automatically control and optimize resource use by leveraging a metering capability at a level of abstraction appropriate to the type of service.

Service Models Service layer Saa. S Paa. S Iaa. S • Software-as-a-Service • Platform-as-a-Service • Infrastructure-as-a. Service Abstraction layer Hardware & Infrastructure 4

Service Models We manage You manage Iaa. S Paa. S Saa. S

Compute Nodes • 7 Dual Twin Olidata (4 hosts) • Host: 2 x AMD 6320 MHz 64 GB RAM • 3 twins dedicated to BESIII • 2 KHS 06 dedicated 175 HS 06 • +1 DG 1 fundings, in test to BESIII + the partial use of 0. 7 KHS 06 from DG 1 Storage • 400 TB gross DELL MD 3660 f + expansion 20 TB net for BESIII 6

Cloud Infrastructure BOSS running Not all random trigger data Reconstruction for phase data Analysis if we have the dst Now: 1) All the main services are installed 2) Direct submission to CE 3) CVMFS tested 4) Job submission in Dirac included 5) Submission from IHEP tested 6) Cloud monitoring: Zabbix 7

Test Cloud – Cpu and Mem cpu load mem load 8

Test Cloud – Cpu and Mem cpu load mem load 9

Cloud – Open. Nebula Interface 10

Cloud – Zabbix Monitoring mem load cpu load 11

Cloud – Running Jobs 12

Cloud – IHEP Job Submission 13

BESIII Cloud Lab

Cloud – Future Plans • Dirac server and client configuration • Storage elements if new storage is funded • Multi-purpose cloud monitoring • Integration with other clouds • Jobs user friendly • Multiple or multipurpose OS • Priorities jobs in the same experiment jobs in shared resources • Elastic instantiation of VMs 15