The CX4 range replaced the CX3 and looking at available cache levels, for comparable positioned arrays, we increased the useable write cache e.g. CX3-10 to CX4-120, CX3-20 to CX4-240 and so on. With the CX4 series, we now look at the available write cache on a CX4 to be the maximum amount of cache on that array, not the memory, and then take some allocation for read cache. This differs to the CX3 and previous generations as we have greater write cache availability and increased levels supported on the CX4 i.e. some ¿older¿ models would have more total cache but had restrictions on how much could be used for write cache.

With the CX4 several factors contributed to the increased memory required by the environment. These include the move to a 64 bit OS, the increased drive count supported by each array model, the increased replication counts supported by the array, and the greater granularity for some of the replication products that results in great improvements to re-sync operations for SnapView clones and MirrorView LUNs.

From the performance testing across the CX4 range the scaling data we have observed for random operations indicates the available cache levels are sufficient to provide a good balance of performance with a level of predictability in line with the disk architectures used. We also have tuning options for environments that may benefit from tuning. Options include cache page size, watermark settings, global allocation of cache, read/write cache enable per LUN, fixed and variable pre-fetch per LUN. In the majority of deployments these optimization adjustments are not necessary and the default options are suitably used.

~Keith