I have done this in the past:: pin "user" (T30 for me) table, this will drop IO to database.. have see this alot..
On Thu, Sep 8, 2011 at 2:36 PM, Guillaume Rheault <guilla...@dcshq.com>wrote: > ** > Hi Joe, > > I got to disagree with you again... but I guess this is what makes this ARS > list fun! > > Pinning a table into memory is not an overkill, it is quite simple to do, > you can ask your Oracle DBA. Since the cost of physical memory is lower and > lower every year, it is actually more cost-effective to add some more memory > to your database server and pin look-up tables, than optimizing the searches > to these look-up tables; optimizing the searches will involves one or more > of the following: > > - Possible DBA time to analyze the performance of queries > - Remedy Admin/Developer/Consultant time to figure where those sub-optimal > searches are being issued from, and modify them > - Possible customizations to the ITSM application (which is what everybody > is trying to avoid) > > Pinning a table into memory involves: > > - Small amount of DBA time to alter the T table to pin it. > - Small amount of sys admin to add memory in the database server (this cost > is a one time cost) > > See, when you pin the table in memory, it does NOT matter if your queries > are crappy or inefficient, since the table data is in memory; that's the > beauty of it! > > While you are at it, you may as well pin the T table related to the User > form. > > cheers, Guillaume > > > ------------------------------ > *From:* Action Request System discussion list(ARSList) [ > arslist@ARSLIST.ORG] on behalf of Joe Martin D'Souza [jdso...@shyle.net] > *Sent:* Thursday, September 08, 2011 12:33 PM > > *To:* arslist@ARSLIST.ORG > *Subject:* Re: Performance question CTM:People timing > > ** > > Yes I agree you would want to avoid pinning a table to memory whose > contents are changed continuously by way of modifications or additions.. > This would result in frequent memory writes which would beat the purpose of > why you choose to pin it to memory in the first place. > > While the CTM:People table is a good candidate as its contents change less > frequently in most standard environments, unless it’s a B2C environment > where you maintain your customer base in your CTM:People form, if the table > size is as small as 140K, just optimizing searches on it is more than > enough, and pinning it to memory is an overkill.. Optimizing searches on > this table when records are about that much or even upto half a million, > would return the search in less than a fraction of a second anyways.. > > Joe > > > *From:* Guillaume Rheault <guilla...@dcshq.com> > *Sent:* Thursday, September 08, 2011 12:15 PM > *Newsgroups:* public.remedy.arsystem.general > *To:* arslist@ARSLIST.ORG > *Subject:* Re: Performance question CTM:People timing > > ** > Joe, > > well, I disagree with your rationale... actually because it is not a large > table, you can pin in it memory. > Generally speaking, you only pin into memory look-up tables that are used > heavily, and the people form/table is a good candidate. > You definitely do not want to pin a transactional table (like the incident > form). > > Guillaume > > > ------------------------------ > *From:* Action Request System discussion list(ARSList) [ > arslist@ARSLIST.ORG] on behalf of Joe Martin D'Souza [jdso...@shyle.net] > *Sent:* Thursday, September 01, 2011 2:19 PM > *To:* arslist@ARSLIST.ORG > *Subject:* Re: Performance question CTM:People timing > > ** > > For only 140K records I don’t think you need to do anything out of the > ordinary to boost up performance. If your statistics were not updated, it > does make sense as Oracle didn’t know it had to use indexes and was perhaps > attempting table scans assuming the table has no records if the statistics > information it had for row count was 0 or thereabouts prior to updating it.. > > Personally I don’t really think you can consider CTM:People with around 140 > K records to be a large object. Its big but not that big enough to be > considered to pin to memory.. > > Joe > > *From:* John Sundberg <john.sundb...@kineticdata.com> > *Sent:* Thursday, September 01, 2011 1:31 PM > *Newsgroups:* public.remedy.arsystem.general > *To:* arslist@ARSLIST.ORG > *Subject:* Re: Performance question CTM:People timing > > ** True... good suggestion. > > > > > Fundamentally - I was looking for what is "normal" -- what we were seeing > was what we thought was slow. But - just cause you think something is slow - > does not mean that it is slow. Sometimes -- you have to look to your > neighbors and compare. > > > So - thanks to all that shared their timings and system info. > > > > > > -John > > > > On Sep 1, 2011, at 8:30 AM, Guillaume Rheault wrote: > > ** > One more way to make things even faster in Oracle is to "pin" the > underlying T table into memory. > Ask the DBA over there to do that > > -Guilalume > > > ------------------------------ > *From:* Action Request System discussion list(ARSList) [ > arslist@ARSLIST.ORG] on behalf of John Sundberg [ > john.sundb...@kineticdata.com] > *Sent:* Thursday, August 25, 2011 7:25 AM > *To:* arslist@ARSLIST.ORG > *Subject:* Re: Performance question CTM:People timing > > ARSlist: "Where the Answers Are"_ > _attend WWRUG11 www.wwrug.com ARSlist: "Where the Answers Are"_ > -- Patrick Zandi _______________________________________________________________________________ UNSUBSCRIBE or access ARSlist Archives at www.arslist.org attend wwrug11 www.wwrug.com ARSList: "Where the Answers Are"