RE: IDS 9.40 migration-question

snagnews@freenet.de wrote > we would migrate from our old Informix-Online 5.1 to IDS 9.40 > and have probed this on our testserver (Alpha-DS15, 1 CPU, 2 > GB RAM, 70 GB HDD in RAID 10). > > The data was moved from the production-server to the > test-server with dbexport (option -ss is unknown in 5.1) and > dbimport databasename -d dbspace. > > We have a few questions about our migration : > > 1.) we have two databases in the informixserver, should every > database use an separate dbspace? > > 2.) we have defined one tempdbs with appr. 1,4 GB for our > databases with 22 GB Data in it, is this ok for an OLTP-System? > > 3.) should we create all indicies in a separate dbspace? > > 4.) are big-chunks better than 2GB-chunks (because we have > already created 2GB-Volumes over the Tru64 5.1B - LSM volumemanager) One question should be "do I need to use LSM for striping and mirroring or can I just use Informix mirroring and split my drives over several dbspaces avoiding the need for LSM all together" > > 5.) if we put the 300 GB rootdbs in the first chunk, and all > chunks reside on the same raid-box, is there a reason not to > put physdbs and logsdbs on the same chunk with defining the offset? You have gone from 70GB to 300 GB ? How many disks ? I would create multiple dbspaces and have physbs on a separate spindle from rootdbs. > > 6.) Should the rootdbs on a volume over a volumemanager begin > on offset 100? If you are using LSM that takes care of the offset, if using raw disks directly you do need the offset. Colin Bull _______________________________________________________________________ This email has been scanned for all known viruses by the MessageLabs Email Security System. _______________________________________________________________________ sending to informix-list

> > > > 4.) are big-chunks better than 2GB-chunks (because we have > > already created 2GB-Volumes over the Tru64 5.1B - LSM volumemanager) > > One question should be "do I need to use LSM for striping and mirroring > or can I just use Informix mirroring and split my drives over several > dbspaces avoiding the need for LSM all together" I prefer the use of hardware raid by raid-controller, we have an MSA30-Box with Raid-Controller and will not use LSM for striping/mirroring. The only thing is, that LSM is able to split large disks into several chunk-volumes of 2 GB (was needed by Informix < 9.xx) > > > > 5.) if we put the 300 GB rootdbs in the first chunk, and all > > chunks reside on the same raid-box, is there a reason not to > > put physdbs and logsdbs on the same chunk with defining the offset? > You have gone from 70GB to 300 GB ? How many disks ? I would create > multiple dbspaces and have physbs on a separate spindle from rootdbs. Sorry, i've made a mistake, the rootdbs is 300 MB, not 300 GB. On the two disk-boxes, we have hardware-striped and mirrored 16 disks with 36 GB and 1 hot spare disk. This disk we have partitioned in 3 partitions, one for the production-database, one for test purposes and one for our arch2disk with filesystem. The database-partitions we have encapsulated as LSM disks and these db-partitions we have splitted in 2GB volumes. All runs on the same physical hardware. This is the hardware, we use. > > > > 6.) Should the rootdbs on a volume over a volumemanager begin > > on offset 100? > > If you are using LSM that takes care of the offset, if using raw disks > directly you do need the offset. We think LSM takes care of the offset. But we don't really know it. In which way we can do a check that the rootdbs is damaged, if we use the offset of 0. Andy > > Colin Bull >