Read Rate Scalability Clause Samples

Read Rate Scalability. This tests the performance of the storage systems under concurrent reads. It is essentially similar to 5.2; however in this case, multiple hosts are recommended as targets to distribute the load across them in order to ensure we are looking at the performance of the storage system and not the client node. For this test, scripts should be used to select a random file from the last run of the 5.2 tests as input, and copied from the storage system to local disk on the client node. These tests will again be run for a single file then in step of 5 concurrent copies out of the storage system (i.e. with parallelism of 1,5,10,15,20, ) until the rate out reaches 200% of the single file rate, with the single file transfer being repeated for a random file 5 times; randomisation is important to ensure that the file is not held in memory by the storage system after the first transfer. If time permits, this test should be repeated similarly for each step, but if time constraints limit accessibility then a single run will at least give some indication of scalability. As before, sites should monitor the longest time for any transfer. If sites have monitoring in place to look at the network throughput to the storage system, this should also be recorded for each step. Files should not be deleted following these tests since they are required for subsequent tests.