Hi, I have a suggestion for improving primo cache.
It would be nice if there was a tool to analyse the block usage of the cache.
Let’s say you are using 128KB block size, if we could calculate how much waste on average per block.
So for example the average block size being cached is 8K in size, the waste in average would be 120kb per block, the user can then re-configure his block size to something more appropriate.
It’s not necessarily easy to find the optimal block size and a tool to figure this out would be very appreciated.
Of course you would need the cache to fill out and stabilize to have an idea of what it’s actually using.
Thank you
Cache block analysis
Re: Cache block analysis
As far as I understand it, this larger block size can actually work to your advantage because if you read 8 Kb, in a way it performs the act of "prefetching" the other 120 Kb of data around that 8 Kb block. (Someone correct me if I am wrong but I did read elsewhere that this is how it works).
Full disclosure is that I am involved with the creation of defragmentation software, UltimateDefrag, and there is a way to use this way PrimoCache handles the blocksize to your advantage if you use UltimateDefrag in that UD enables you to specifically put the files specified in Windows layout.ini to the beginning of your hard drive. So that during boot up, since most of your system files will be grouped together on your hard drive it is automatically going to conveniently read some boot time data into L1 before the data was even requested because it will have been preread into cache as a result of 128 Kb block size.
Full disclosure is that I am involved with the creation of defragmentation software, UltimateDefrag, and there is a way to use this way PrimoCache handles the blocksize to your advantage if you use UltimateDefrag in that UD enables you to specifically put the files specified in Windows layout.ini to the beginning of your hard drive. So that during boot up, since most of your system files will be grouped together on your hard drive it is automatically going to conveniently read some boot time data into L1 before the data was even requested because it will have been preread into cache as a result of 128 Kb block size.
Re: Cache block analysis
Sorry, but I'm afraid that this function will not be implemented. It affects the performance quite a lot to trace the block utilization ratio. Besides, PrimoCache might auto fill the whole cache block, considering the read-ahead, prefetching, populating L2, etc.