Re: Initial configuration questions...

jan, > two 36G scsi arrays. The system will need to support two applications: an > online transaction system (very light load - only a few operators, a couple of > gigs of data at most), and a heavy-duty research type database (50-60 gigs, > nightly batch updates only). there is a lot of information that is needed here to put all this in perspective, but it sounds to me like you will be running both an oltp and a dss system. i personally don't like the idea of two instances running on the same machine in a production environemnt, but it is doable. basically, if the load is light on the oltp stuff, and a lot of the dss stuff takes place in batch at night, then these can certainly work well under the same system. the last system i set up was a sun with a storage array, and was the other way around. it was a heavy oltp system, and light dss, mostly during the evenings. so during the day we had the server running with no parallel processing allowed, and at night kicked it in full bore. the system worked well that way. > 1. Should we run one or two server instances? The research app is basically a > data store only, and does not need transactions. you control transactions, (logging), on a database by database basis, so you turn logging on for the databases that need it, and you don't turn it on for the ones that don't. > 2. We'd like to use Informix on raw devices - should both apps share the same > dbspace? The research app consists of a potentially large number of tables - > will we run into table naming problems (does Informix use namespaces?) you should definately use raw devices! you don't worry about them sharing the same devices, that is at the informix level. with my setup, i had just about everything going at once, my important dbspaces were set up with raid-5 partitions on the storage array, the critical dbspaces, (root dbspace, log dbspaces) were also mirrored, and then raid 0, or basic striping for the rest of the dbspaces. table naming takes place within each database. so two databases can have the same table names, because they are different entities to the database. it's like you can have two files with the same name in two different directories. > 3. Will we be better off using Solaris striping, or just chunking the drives > "by hand" to achieve some level of striping? see above. > I would appreciate any comments/suggestions from the trenches... what you are getting into is really quite involved, and you need a good understanding of the informix system to set it up correctly. i would suggest you or whoever is going to set this thing up go for some informix training, and/or find someone knowlegable to help you out. it will be really easy to implement a system like this and have it perform like a pig if you don't know the details of how to set a system like this up. hope that helps. good luck! mickm