And my point is that Windows has no global option to turn off its caching and do directio, so you must have misinterpreted the questions in the other posts. ATTO does directio at the application level by explicitly making direct io calls instead of buffered io calls. You can't do it system wide or on a disk basis. An application has to choose not to do buffered io.
So if you disagree, explain to me how you would turn on/off the Windows caching.
First of all, the performance tests include the write scores as well as the read scores so discussion of whether or not to turn off the windows disk cache is important as regards to the 'write' portion of the situation and the actual throughput on the device.
HDD and/or SDD efficiency IS affected by ones decision to turn off the write cache. Notice that it is fairly common to refer to the windows cache as 'the windows disk cache' which is NOT referring to any on board physical disk cache.
Second, in referring to 'the controller' that refers to the motherboard controller whether it is part of the 'south bridge' as in the new AMD boards or a distinct controller chip mounted on the motherboard.
You seem to be so focused on a physical disk cache or the 'controller' within a SSD that you are not even considering the existance of any other type of controller or cache. Also I might add that even though SSDs do have an onboard controller, that onboard controller works in conjunction with the motherboards own onboard controller whether or not that 'motherboard controller' is part of the south bridge, an individual chip attached to either the north or south bridge, or a distinct controller add in card.
My own motherboard has a built in controller as part of the AMD 'south bridge', a discrete Marvell controller chip attached to the south bridge and finally yet another Marvell discrete chip attached to the north bridge.
An add in controller card may also contain it's own cache to speed things ups, however, I never referred to any cache other than the Windows 'system' or 'disk cache'. I also never referred to any controller other than the motherboards built in controllers, and specifically NEVER referred to or impled a physical HDD or SDD onboard controller.
As evident in the following thread: Windows write cache on vs off, Vertex-3 120 [Archive] - OCZ Forum there is much confusion as to which cache is being referred to.
Rest assured that my references to using or not using the cache does refer to the RAM cache that windows uses whether it is CALLED 'windows cache', Windows system cache', 'systems cache' or even 'windows disk cache'.
Indeed, the term 'Windows Disk Cache' came about early in the process and only refers to the fact that it was the disk drive which was being cached.
I guess that I should also respond to your post about the cache only containing 'files' in that the cache contains not only 'files' being read, but also files being written out, This is because it is common for something being written out to be immediately being read back in shortly afterwards. By keeping whatever was written out in the cache, it is readily available if needed and avoids more disk access. Also, 'files' isn't restricted to complete files but may also be portions of a 'file' such as a data base where only a small portion is being accessed.
I think that I have more than covered the subject involved and therefor I won't be responding any further to this thread.
You haven't told me how one would go about globally, or per disk, turning off the windows caching, like ATTO did for its benchmark application. It is because you cannot. Prove me wrong and tell me how or quit blathering on.
1. You posted ATTO scores with directio disabled
2. You said this was to show some posters how important enabling write cache is (windows refers to on-board disk cache as "write cache") but you say you meant windows caching by this. That is OK.
Now I ask you, how one even disables windows caching globally that you would even need to enable it and you cannot answer me.
Last edited by GeneO; 16 Nov 2011 at 19:22.
Quote