Re: Re: TWO questions
Encuenta = To take care of. (+/-) So the questions may read as: What Do I need to do to use fragmentation/mirroring with Informix? Also advise on Informix parameters is required. Chucho! -----Original Message----- From: Jonathan Leffler <jleffler@earthlink.net> To: informix-list@iiug.org Date: Sun, 14 Nov 2004 05:31:39 GMT Subject: Re: TWO questions Dooku Maul wrote: > I have two questions fererring to fragmentation and mirroring > 1. FRAGMENTATION > That criteria are used to make a fragmentation and as they are the > aspects of informix to have encuenta. > > 2. MIRRORING > As they are the parametros to take encuenta for the configuration and > boot from mirroring. > > My Platform: > HP-UX 11.11, with Informix 7.31 FD6, with 4 CPU's and 12 GB What does 'encuenta' mean? It isn't recognized by babelfish.altavista.com as either Spanish or Portugese. And, unfortunately, your question doesn't make much sense without knowing roughly what 'encuenta' means. Regarding Fragmentation: Informix supports pretty flexible range expressions. Regarding Mirroring: Informix does take some mirror information from the ONCONFIG file, but the rest is stored in the 'reserved pages' for the instance. Those are not very helpful comments - but I don't understand the question well enough to do better. -- Jonathan Leffler #include <disclaimer.h> Email: jleffler@earthlink.net, jleffler@us.ibm.com Guardian of DBD::Informix v2003.04 -- http://dbi.perl.org/ Jean Sagi jeansagi@myrealbox.com jeansagi@yahoo.com sending to informix-list
Jean Sagi wrote: > Encuenta = To take care of. (+/-) > > So the questions may read as: > > What Do I need to do to use fragmentation/mirroring with Informix? > > Also advise on Informix parameters is required. Thanks for the assist. At least the questions now make sense. To use fragmentation, there are a couple of necessary pre-requisites. The first (in currently available versions of IDS) is to have a number of dbspaces - at least one dbspace for each fragment of the table which will have the most fragments. (That limitation should go in the next version of IDS.) The other pre-requisite is a plan for how each table will be fragmented. There are endless possible variations for this. The 'simplest' is to use ROUND ROBIN fragmentation, but that is seldom the most sensible choice. Usually, there is one attribute, sometimes several attributes, that can be used to partition the data into separate ranges - range partitioning. This is usually beneficial because the optimizer may be able to avoid scanning whole fragments (fragment elimination) if the queries are based on expressions related to the fragmentation expressions. The classic example is something like monthly sales figures. You might fragment the table into 12 or more fragments, one per month. Then, if you are looking for just the current month's data, or the last completed month's data, you only need to scan one of twelve dbspaces. Further, it is easy to attach and detach fragments, so you can roll the old data out of the system quite easily - and create new fragments each month for the new data. Other examples of range fragmentation are semi-geographic - breaking up tables of addresses based on the postal code (zip code in the USA). Once you have your planning in place, the only implementation steps are running the ALTER TABLE statement to initialize the fragmentation scheme - and ensuring appropriate backups are done. With mirroring, there are multiple levels at which mirroring can occur. If you can use hardware mirroring - implemented by your RAID system, or the LVM, or whatever - then you should probably use that in preference to IDS mirroring (because the hardware mirroring only requires one write system call to copy to both copies of the data to disk - whereas if IDS does the mirroring, it requires two write system calls, one for each copy). If you decide to use IDS mirroring, you need to enable it in the ONCONFIG file (I believe), and then create mirrored chunks for the dbspaces where you want IDS to mirror the data. Thereafter, IDS will write each page to each of the two (mirrored) chunks. If you use IDS mirroring, mirror your critical dbspaces first - rootdbs, logical log, physical log, key data tables. (Even if you use hardware mirroring, you still mirror your critical dbspaces first, of course - the same set of suspects.) There's a lot that could be said, no doubt. With luck, my explanation is simple enough to understand. > -----Original Message----- > From: Jonathan Leffler <jleffler@earthlink.net> > > Dooku Maul wrote: >> I have two questions fererring to fragmentation and mirroring >>1. FRAGMENTATION >> That criteria are used to make a fragmentation and as they are the >>aspects of informix to have encuenta. >> >>2. MIRRORING >> As they are the parametros to take encuenta for the configuration and >>boot from mirroring. >> >>My Platform: >>HP-UX 11.11, with Informix 7.31 FD6, with 4 CPU's and 12 GB > > What does 'encuenta' mean? It isn't recognized by > babelfish.altavista.com as either Spanish or Portugese. And, > unfortunately, your question doesn't make much sense without > knowing roughly what 'encuenta' means. [...] -- Jonathan Leffler #include <disclaimer.h> Email: jleffler@earthlink.net, jleffler@us.ibm.com Guardian of DBD::Informix v2003.04 -- http://dbi.perl.org/