Re: Online Instance Question

>From: pearlkoh <pearlkoh@singnet.com.sg> >Date: Mon, 10 Feb 1997 22:57:54 +0800 >X-Informix-List-Id: <news.33708> > >If on a single host machine, I have a large database (A) which is >constantly being updated and other databases of similar size or >relatively smaller sized databases resident, which would be a better >Informix solution, given the following: > o Informix environment - Informix 5.03xxxxx > o Database A is not dependent on any other database. > o The application that owns database A is does not access any other > database. > o The applications that own the other databases, do need to access > database A. > >I guess the two possible Informix db solutions would be: > 1 Create a large online instance and lump all the databases in it. > or > 2 Create online instances such that databases that logically belong > together reside in their own online instances. Access to "external" > databases would then be made through I-STAR. Both solutions are workable. >Given that I am running all the databases on one host, would choosing 1 >over 2, imply that I have lower network overheads? Using solution 1 would give lower network overheads; there is no networking required (other than for clients talking to the servers, perhaps). >Does accessing a "server" database residing on the same machine, require >requests to go out into the network or does loopback confine it within >the machine? I think, but don't know, that loopbacks do not actually send the data onto the physical network. >Are there any speed advantages by choosing option 1? The advantage of option 1 is that you probably make better use of the machine resources. For example, assume you have, 64 MB of memory in your machine and that you can afford to allocate 16 MB of that to OnLine. With option 1, you can simply allocate it all to the one OnLine system, and it will be used as required. If database A is in heavy use, but the others are passive, then all that memory is available to database A. When the other databases are in use, a little less will be available to the users of database A, but it is all handled dynamically. By contrast, if you have multiple OnLines running, then you need to determine a static split -- maybe 12 MB for database A and 4 MB for the others. When the other databases are inactive, then that 4 MB is not available to database A. On the other hand, using the multiple OnLine systems guarantees the other databases a certain amount of space, regardless of the contention for database A, so it depends on what is most important to you. Yours, Jonathan Leffler (johnl@informix.com) #include <disclaimer.h>