Dr Jose Antonio Coarasa Perez (Department of Physics - Univ. of California at San Diego (UCSD))
The Open Science Grid middleware stack has seen intensive development over the past years and has become more and more mature, as increasing numbers of sites have been successfully added to the infrastructure. Considerable effort has been put into consolidating this infrastructure and enabling it to provide a high degree of scalability, reliability and usability. A thorough evaluation of its performance and scalability is important in order to assess the effective deployment of new sites and to compare new versions of the middleware to existing ones. To do such scalability tests, requires testbeds comparative in size with the largest available sites on OSG. In this paper we describe testing procedures, available testbeds, and tools to test and monitor the scalability and performance of a deployed site. We use these methods to review the scalability and performance of a deployed Computing Element (CE) and a deployed User Interface (UI) giving independent measurements for them. The limiting pieces of the infrastructure are singled out especially in the context of multicore/multiprocessor hardware. Comparison of the different versions of the infrastructure is given.