Use of glide-ins in CMS for production and analysis

Abstract

On behalf of the CMS Offline and Computing Projects)Abstract. With the evolution of various grid federations, the Condor glide-ins represent a key feature in providing a homogeneous pool of resources using late-binding technology. The CMS collaboration uses the glide-in based Workload Management System, glideinWMS, for production (ProdAgent) and distributed analysis (CRAB) of the data. The Condor glide-in daemons traverse to the worker nodes, submitted via Condor-G. Once activated, they preserve the Master-Worker relationships, with the worker first validating the execution environment on the worker node before pulling the jobs sequentially until the expiry of their lifetimes. The combination of late-binding and validation significantly reduces the overall failure rate visible to CMS physicists. We discuss the extensive use of the glideinWMS since the computing challenge, CCRC-08, in order to prepare for the forthcoming LHC data-taking period. The key features essential to the success of large-scale production and analysis on CMS resources across major grid federations, including EGEE, OSG and NorduGrid are outlined. Use of glide-ins via the CRAB server mechanism and ProdAgent, as well as first hand experience of using the next generation CREAM computing element within the CMS framework is discussed.

GWpilot: Enabling multi-level scheduling in distributed infrastructures with GridWay and pilot jobs

“…However, overheads are actually appreciable with task duration below five minutes. Results obtained in other cases seem to be comparable with those of a FRFS scheduling approach [76], or even in [52], as shown in Figures 5-(d, e, f). This is indicative of the design and implementation feasibility of GWpilot.…”

“…Makespan obtained with every pilot system is listed in Table 6, which must be provided as the principal speedup measurement from a user point of view. However, to allow a proper comparison among systems and against other approaches, other performance metrics used in previous works [47,76,44,52] are considered. Thus, in addition to the final makespan parameter, Table 6 compiles the number of pilots submitted and effectively enrolled as a measurement of the efficiency in provisioning.…”

“…With this system, great results were obtained [52]. However, the CMS community has two PMS [18], one for data analysis [53] and one for Monte Carlo production [54], that can submit tasks to Condor-G, gLite/UMD WMS, ARC [55] and some LRMS through a connector called BOSSLite [56].…”

“…In relation to provisioning, the number of effectively enrolled pilots presented in Tables 6 and 9 is also important. There were tests with more than 9,000 successfully enrolled pilots [44], achieving maxima of above 200 pilots per minute [52]. However, it is interesting to provide the average number of enrolled pilots and the percentage of discarded pilots per hour [81], which are between 200-300 and above 5%, respectively.…”

Towards automating the configuration of a distributed storage system

“…One of the many instances [1] [2] [3] [5] [6] [15] [16] where this technique is used, yet relevant to Grid settings, is storage system 'glideins' [1] [18] [19] [20]. In this scenario, the components of a storage system are submitted together with a batch application and the storage system is instantiated on the fly to aggregate the storage resources available on the nodes allocated to the application.…”