We have a box that is file server that appears to be having some I/O issues so I have been reading up on tuning for improved I/O performance. RedHat has an article regarding how to tune your VM - <a href="http://www.redhat.com/magazine/001nov04/features/vm/#tuning-vm">http://www.redhat.com/magazine/001nov04/features/vm/#tuning-vm</a> Does anybody have any experience tuning bdflush? The docs I am reading appear to be contradicting, or atleast they appear that way to me. Basically there are 6 tunable parameters for bdflush – nfract, ndirty, interval, age_buffer, nfract_sync, & nfract_stop_bdflush. The docs state their definitions to be <code>nfract</code><code> -</code> The percentage of dirty pages in the buffer cache required to activate the <code>bdflush</code> task <code>ndirty</code> - The maximum number of dirty pages in the buffer cache to write to disk in each <code>bdflush</code> execution <code>interval</code> - The number of jiffies (10ms periods) to delay between <code>bdflush</code> iterations <code>age_buffer</code> - The time for a normal buffer to age before it is considered for flushing back to disk <code>nfract_sync</code> - The percentage of dirty pages in the buffer cache required to cause the tasks which are writing pages of memory to begin writing those pages to disk instead <code>nfract_stop_bdflush</code> - The percentage of dirty pages in buffer cache required to allow <code>bdflush</code> to return to idle state It then says Generally, systems that require more free memory for application allocation want to set the <code>bdflush</code> values higher (except for the <code>age_buffer</code>, which would be moved lower), so that file data is sent to disk more frequently and in greater volume, thus freeing up pages of RAM for application use. This, of course, comes at the expense of CPU cycles because the system processor spends more time moving data to disk and less time running applications. Conversely, systems which are required to perform large amounts of I/O would want to do the opposite to these values, allowing more RAM to be used to cache disk file so that file access is faster. To me, this ^ looks like the opposite of what the definitions for each parameter state. I would think if I want to allow more RAM to be used for faster disk access than I would want to INCREASE all of the parameters ( I am not positive about nfract_stop_bdflush though). Am I reading this wrong because to me it looks like it is contradicting itself. For example, if I want to free up RAM why wouldn't I lower nfract, thus lowering the required amount of dirty pages in the buffer cache before it writes back to disk? Earlier in the article they have an example that states <tt>#bump up the percentage of memory in use to activate bdflush</tt><tt> vm.bdflush="40 500 0 0 500 3000 60 20 0"</tt> If I take that and compare it to the standard RHEL 3 /proc/sys/vm/bdflush of 50 500 0 0 500 3000 80 50 0 It would seem to me that this would lower the amount of memory in use to activate bdflush. Am I missing something and just completely not understanding this or is this doc and example incorrect? Thanks, Nick