Performance Monitoring

One of the key issues in the LHC Computing Grid is the performance of individual jobs as well as overall throughput.. When we speak about such issues, we think about the performance of a single application, of a library as well as of a function or an entire stack of libraries, which in all cases are dependent on the underlying operating system and hardware. In order to optimize the software, first we have to understand how the application runs. This is where performance monitoring comes in. In openlab, we collaborate with our openlab partners on monitoring solutions. We take advantage of performance monitors which are available in modern CPUs. We use different interfaces, libraries and tools in order to access performance resources.  A particular case of performance monitoring is profiling were we are heavily involved in collaboration with programmers and physicists from the LHC experiments.