The aim of this on-going work is to study the scalability of the back-end network of storage sub-systems in terms of the number of disks that can be linked to the network. It is well known that without considering the limitation of back-end network, increasing the number of disks in a RAID based storage system will increase the parallelism, and so can lead to a higher performance. Moreover, to save money on the back-end network, it is common practice to scale the number of disks rather than the number of independent access pathways. However, in a real system there is a limitation on the scale of storage sub-systems (controller cache size and number of disks that can be included in one system) due to the limitation of interconnection network. This is because the back-end interconnection networks are shared by all the disks and the RAID controllers in a storage sub-system. The more disks are added to the system, the higher the contention for the shared media. When the number of disks and cache size in a RAID system reaches a certain threshold, there will be no further gain in performance by adding more disk or cache due to the saturation of the back-end network. Therefore, in order to design a scalable storage sub-system it is critical to study the saturation characteristics and scalability of the back-end network. Previous work has focussed on sequential accesses only when working out when the back-end network becomes saturated, this does not represent a 'normal' workload. This work uses a workload based on the Storage Performance Council SPC-1 benchmark and so uses a representative loading.
|Title of host publication||Proceedings of the 5th USENIX Conference on File and Storage Technologies|
|Place of Publication||Berkeley, CA, USA|
|Number of pages||1|
|Publication status||Published - 2007|