Re: DSA follies

In article <5jjr7f$gjp@cssun.mathcs.emory.edu>, David Coburn <David_Coburn@wvccg.wvus.org> writes > > > > >Hi. I've been fighting now for two months with a problem. I'm hoping >that somebody else has run into a similar problem and has a >workaround. > >I have an ESQL/C program that declares a cursor with hold and then >performs an inner loop to do some updates to various other tables. >The problem is that beginning with a migration to an HP 9000 K420 with >4 CPUs, I find that the query on the outer hold cursor isn't being >parallelized, where it once was on the older Unisys PTX machine. The >HP has Informix 7.22.UC1, the Unisys 7.10.UC5. Frankly, this old >Unisys machine, several generations out of date (the 4 CPUs are 33 >Mhz. Pentium chips, for example) is much faster on these types of >queries than the new HP. I have checked, and found that (a) >PDQPRIORITY is being set (verified via onstat -g mgm); (b) the >optimizer is allocating a correct number of threads for my query (in >this case, 5, again verified by onstat -g mgm), and finally, onsat -u >reports that there is (guess what?) only a single thread being >allocated for this query. On the old PTX box, once again, it >allocated sufficient threads, verified by the same onstat outputs as >shown here. Correct - hold cursor declared with hold are not parallelized... Several other types 'queries' are not parallelized - can't remember which... > >Support is now stating that this is not a bug, but a >feature...apparantly parallelizing a hold cursor is "too dangerous." >Besides, there is no version of DSA that permitted this to be >parallelized in such a fashion, according to them. When I pointed out >that the 7.10 on PTX did, it became "oh, then try setting your >isolation level to..." Been there, done that, no luck. Now I'm off >on yet another support engineer, but I get the impression that these >poor guys are having to toe the company line here without giving their >true feelings in the matter. John and Andy are both fine engineers, >and this is NOT a slam on them, but...I'm stuck in the middle. > >Has anybody else run into this? I can reproduce at will on my HP, but >am tired of having to support two database machines for one >database...the production jobs overload the capacity of the HP (and >there aren't too many, folks!) so I have to run the two in sync. This >wouldn't be too bad, but Informix isn't paying for the upkeep of this >old box, and it is on its last legs with no maintenance support >available. Sigh. > >Thanks, > >David Coburn >Sr. DBA >World Vision > -- David Williams