Re: Is there a benefit in having indexing on separate spindles

In article <54k48a$9h2@cssun.mathcs.emory.edu>, Peter Tashkoff <TASHKOP@kiwi.co.nz> writes > from Data. > >Hi >I am been speaking recently with an >Oracle DBA (Tired looking fellow). >He mentioned to me that he always made >a practice of storing table indexes on >separate spindles from table data, and >that this provided performance >benefits. >I would be interested in hearing from >anyone who is doing this in informix >and is experiencing performance gains. >Or even peoples opinions on the >subject. >TIA. >-- >-- >Peter Tashkoff <tashkop@kiwi.co.nz> >NZ Kiwifruit Marketing Board Usual >disclaimers apply. >This posting may not be used by any >party to vilify another. > > Depend on how many disks you have available. Definately do not have it all the Online Instance on one disk. Ideally you have the following 1. Under UNIX the Online instance is not on the same disks as /, /tmp or where mail gets stored as these disks will get quite badly hammered. 2. Root dbspace on disk one 3. Logical logs on disk two 4. Physical logs on disk three Note: we are up to at least five disks already. 5. Separate high usage database tables and associated indicies on seperate disks. i.e one disk = one table + associated indicies. Note: assume five large tables now we are up to 9 disks. 6. Fragment each high usage table/index across separate disks. i.e. one disk = one fragment of one table + one fragment of associated indicies. Note: Assume three fragments per table now we are up to 4 + (3*5) = 19 disks. 7. Indicies on seperate disks from table. Notice how fairly rapidly you need lots and lots of small disks! Also by the time you reach step 7. you probably have > 20 disks!! -- David Williams