After the first part on this topic (I/O Block Size – Part 1) I have received some interesting feedback. And now I wanted to make one point clear again here which I think I made in my first part.
So the only interesting point I have received here is the whitepaper link for the HP StorageWorks EVA Configuration best practice. Now on this white paper they emphasizes on the fact that Transfer size influences sequential performance.
Applications such as data warehouse and business intelligence that have a high percentage of sequential write application can improve array performance by verifying that the write transfer size is greater or equal to 32K and is a multiple of 8K (for example, 32K, 40K, 64K, 72K, 80K, … 128K). These transfer sizes simplify the cache management algorithms (which reduce the controller overhead to process a command) and reduce the total number of I/Os required to transfer a given amount of data. Storage systems typically choose these write transfer sizes, but operating systems, file systems, and databases also provide settings to manage the default transfer size
Best practice to improve sequential write performance: Tune the write transfer size to be a multiple of 8K and no greater than 128K.
So in our first article we were sure that it all depends on the workload we are Virtualizing.