Experience in using commercial clouds in CMS

Abstract

Abstract. Historically high energy physics computing has been performed on large purposebuilt computing systems. In the beginning there were single site computing facilities, which evolved into the Worldwide LHC Computing Grid (WLCG) used today. The vast majority of the WLCG resources are used for LHC computing and the resources are scheduled to be continuously used throughout the year. In the last several years there has been an explosion in capacity and capability of commercial and academic computing clouds. Cloud resources are highly virtualized and intended to be able to be flexibly deployed for a variety of computing tasks. There is a growing interest amongst the cloud providers to demonstrate the capability to perform large scale scientific computing. In this presentation we will discuss results from the CMS experiment using the Fermilab HEPCloud Facility, which utilized both local Fermilab resources and Amazon Web Services (AWS). The goal was to work with AWS through a matching grant to demonstrate a sustained scale approximately equal to half of the worldwide processing resources available to CMS. We will discuss the planning and technical challenges involved in organizing the most IO intensive CMS workflows on a large-scale set of virtualized resource provisioned by the Fermilab HEPCloud. We will describe the data handling and data management challenges. Also, we will discuss the economic issues and cost and operational efficiency comparison to our dedicated resources. At the end we will consider the changes in the working model of HEP computing in a domain with the availability of large scale resources scheduled at peak times.

Enabling HPC systems for HEP: the INFN-CINECA Experience

“…This has been chosen in view of further manipulating the payloads, as explained later. Many combinations of such parameters were tested, starting from very simple ones (only low memory / low IO production jobs), to the most demanding (accept full Monte Carlo production StepChains [18], and end-user analyses). The final configuration, established after the commissioning of the 40 Gbit/s reserved link to CNAF, has demonstrated that CINECA can be used as a standard CMS facility, on which almost all the workflows can be executed.…”

Assessment of financial returns on investments in cyberinfrastructure facilities

“…The works cited above generally show that investing in a local facility is more cost-effective than using commercial clouds; it is also possible to assess the financial efficiency of using cloud resources for specific workflows. Bauerdick et al [17] performed an experiment comparing the use of a commercial cloud with workflow-specific optimization to the use of a local facility. Specifically, they compared AWS (Amazon Web Services) spot instances for analysis of CMS experiment data [9] versus the Fermilab HEPCloud [6].…”

“…The comparison showed that it is possible to get costs of commercial cloud services close to the actual costs of local facilities. But, after using over 15 million hours (approximately a month of time on a modern super-computer) they concluded, "The steady-state cost [of AWS] came to 1.4 ± 12% cents per core-hour, which is not much larger than the estimated 0.9 ± 25% cents per core-hour for the Fermilab data center" [17]. Such small differences do not sound like much, but they can accumulate over a project lifetime to a cost difference of hundreds of thousands to millions of dollars.…”

Computing challenges of the CMS experiment

“…An AWS academic 9 to 1 matching grant was awarded to CMS that allowed the purchase of $300.000 of credits for computing, storage, and network charges for an investment of $30.000. The size of the award was based on an estimate of the cost of one month of large-scale processing [2].…”

“…The results are summarized in table 1. The computing performance had reached roughly equivalent value, in terms of costs cloud costs are higher, but approaching equivalence in the future [2].…”