dCache performance and its correlation to the pool size

The size of hard disks is increasing every year, therefore it is natural that dCache [1] pool sizes are growing as well. The determination of the correct pool size is rather difficult. From a management point of view, it is preferable to only have a limited amount of pools. However, less and bigger pools aggravate operation. If one pool becomes unoperational, the complete storage behind this pool is offline. Additionally, restarting of bigger pools is time consuming and causes unnecessary delays. However, as a Tier-1 site like GridKa, which is serving all four LHC [2] experiments, the disk space provided behind a single server has to be optimised to the maximum one pool can provide. This paper analyses the coherence between pool size and the amount of transfer movers to result in highest throughput and less administrative impact.