Re: Calculate optimal value for DS_POOLSIZE

>mohitanchlia@gmail.com >Version: IDS 10 >I am trying to determine the optimal value for DS_POOLSIZE and >DS_HASHSIZE. I looked at the onstat -g dsc. I have following question: > >1. How do I determine the correct size of DS_POOLSIZE. Should I look >at our update stats and see number of unique columns that we run >update statistics on and that would become the value of DS_POOLSIZE.>2. How do I determine correct value of number of Hash Buckets ? How >many entries per bucket is good to have ? >3. When I look at onstat -g dsc I see number of distribution entries, >should this be always lower than DS_POOLSIZE Hi, 1. Use the following process to get DS_POOLSIZE and DS_HASHSIZE -Identify the number of columns that have distributions set for the existing Instance of IDS. -Take this value and multiply it by 1.1 to account for any additional overhead, plus to give a little buffer to reduce overflow possibilities. -Divide this number by the depth you want each bucket to be, typically between 4 and 10. -Find the closest prime number to this generated value that is greater than the generated value. This prime number is what should be set as the DS_HASHSIZE value. -Multiply your new DS_HASHSIZE by the same value you originally divided by in order to obtain your DS_POOLSIZE. -Additional tuning may be required 2. Generally 4 is enough for the dept of the buckets. Think of DS_HASHSIZE as an array, but think of the depth of the buckets as a linked list. 3. Yes. If is it around or > than 75%, you are going to start cleaning the cache, if you are at 100% you not only are tuned too low, but you also are adding entries to the cache faster than you can clean the cache.