Re: Checkpoint duration increase with 7.22.U

On Thu, 22 May 1997, Joe Lumbley wrote: > Saw your posting and was really interested in this section. I have a > client on a HPUX 10.20 system and 200 gig of EMC drives. Does it look > to you that the hardware caching is that much better? Could you give > me some idea as to the system you're running? Are you a DSS app or an OLTP > app? What was your buffers setting prior to taking it down to 1000? Is the > entire database on the EMC? How is the EMC set up? That's what it seemed like. Note that our database, while highly active, is very small ( < 2 gig at most sites ). Highly (completely) OLTP. Our standard hardware would be, say, an IBM J40 8 way, or an HP K260 4 way, that kinda thing. The entire database is on the EMC drives, yes, split over 5 or more disks (not counting logs, root, etc) and mirrored (disk mirroring, not Informix mirroring). With Informix caching, we typically set buffers to give us about 50% real memory usage (total). For our systems, this is often between 128mb and 450mb of buffer cache. A little extreme for the size of our database, but it seems to work for us .. assuming standard drives. We were a little hesitant about going with EMC because our application is most definately processor bound for the most part. The drives did help us a little anyway (5-10% speed increase, ish) and we got another boost when we dropped the buffers way down (call it another 5-6%). Not a huge gain, but nothing to sneeze at. > Supposedly, EMC has a white paper in the works on Informix performance. Have > you seen it? 'fraid not. I'll do some digging around the office, though .. if its out, we probably have one, as we're currently doing EMC testing. -Richard