Fast T setup

[costiles]

Previously when setting up the volume groups for AIX, we used small pp sizes because we have a database application - Universe where we wanted as many people to be able to write to a file simultaneously as poossible and with the software striping, we got a good round robin affect with as small a pp size as 4 mbs. On the fast t - we see that the default is 1024 Mbs per pp. I will have 8 or 10 files that will fit in a pp of that size! That would seem to rule out the good distribution of my data over lots of disks - so that many people could acess the file over lots of disks at the same time. It would seem that I am going to contend for that disk where those 8 or 10 files all reside in the same pp. IS there something in the default setup of the Fast Ts that nullifies this way of doing things - or should the same rules apply - ie - small pps so the file gets spread over more disks? The Fast T is set up with raid 1 plus 0.

 

[DaveL67]

I have tried a number of combinations on the FASTt. My FASTt is setup as RAID5. I compared using smaller disks (12GB) each and using one large disk (250GB). I did not see any speed difference between the two. I opted for the larger disk, since the FASTt will present the disk as one large drive to AIX. I have found with DBAa and software vendors that they have the old-school mentality of not wanting to use RAID-5 and most want to use 12 way striping. Though I don't classify the FASTt as tier 1 disk, it does the job. Especially when you have 15K RPM disks in it.

I am also using Flash Copy and Volume Copy if you need any information.

 

[costiles]

We have found that, in a large database application with lots of users, that our performance is not as good with Raid5 as it is with Raid 1+0 and disks are cheep! Having said that it appears that the primary vehicle for good performance resides in the FAST T setup and that it may not make any difference how we setup the volume group and the logical volumes. Am I correct in that summation? - Ferris? Ferris Bueller? - Anyone?

 

[Breslau]

i've played around with a 200 unit a few yrs ago. the only signifigant improvement i saw in tests was turning off cache mirroring. use at your own risk

 

[mjldba]

I had numerous performance issues with a FastT200 attached to a P630 & I saw my best performance improvement when I disconnected it and kicked it to the curb.

After IBM tried almost every available combination of settings and still could not get it to perform, they tried to sell us more disks (increase the # of spindles) rather than admit it would never perform adequately in the configuration they sold us.