Re: Largest Informix Database size

David Williams wrote: > Of course you could probably scale further if you have the hardware, > if 176 is possible I wonder what the limit is, Tim? > > David Williams David, XPS is unlimited in a cluster context, but limited on each coserver in the same context as a typical SMP set-up, but not as limited as the 7.x engine. XPS enjoys a larger page size header ( 56 byte opposed to 24 byte on 7.x ) and a whole host of other engine specifics that take you beyond the typical SMP box with a lot of drives and CPUS. On a typical SMP box, it doesn't matter how many CPUs and drives, it's still ONE BOX. In XPS we depart from the ONE_BOX paradigm into the realm of multi-boxes, multi-CPUs, multi-drives. XPS is the ability to add the third dimension to your data base. Each physical machine using XPS has whatever the posted limits are for a single server. It abides by the usual limits imposed as if it were not a part of a cluster. This is the essence of shared-nothing MPP, using SMP boxes clustered together. All the rules of SMP apply, but with the added benefit of a high-speed communication bus to transport messaging and data between the servers. In a clustered setting with XPS each coserver has to manage only a subset of a query or other SQL activity of a typical data base. The orientation is difficult to imagine in the abstract of someone writing a story about it such as this one, but with careful fragmentation ( partitioning to you Oraclites ) you can span your data not only wide but deep across as many physical computers as you can imagine and pay for. You begin to understand things like data skew and all the other cool stuff only available with XPS. Oracle cannot do this kind of data base management, and will probably NEVER have this capability. The only other contender that I know of might be DB2, but it too is so far behind XPS it will never get there. Redbrick is an SMP-only production, meaning it will never have the clustering capabilities of XPS, but if you had to make a choice between XPS MPP or Oracle SMP or Informix SMP ( 7.x ) Redbrick ain't bad for a single server SMP solution ( from what I've always heard ) if you can't afford XPS. Personally I think everyone should get their hands on XPS, and make clustering as common as the household toaster. XPS has been ported to Linux, but Informix ain't letting it out for some strange reason. They ported it to NT, but nobody could take it serious on NT is my understanding. Imagine offering XPS as a low cost cluster on Linux. To elevate to the realm of clustering, one need only apply the same constructs to physical machines that we do to single physical machines with multi-CPU systems. XPS is the equivalent in physical machine clustering as the multi-CPU ( SMP ) box is to a single CPU machine. It is *that* profound, and thusly, to answer your question, is there a limit with XPS? Only your budget! Tim -- . .- .-- .--- .---- Tim Schaefer .----- tschaefe@bellsouth.net .---- http://www.inxutil.com .--- http://www.datad.com .-- .- .

Tim Schaefer wrote: > other cool stuff only available with XPS. Oracle cannot do this kind of data > base management, and will probably NEVER have this capability. The only other > contender that I know of might be DB2, but it too is so far behind XPS it will > never get there. > Oracle is so far behind these products it isn't even > in the same category. I think Oracle is at least 10 years behind Informix > now in technology and engine design. The latest SQL-Server from Microsoft > is probably more advanced than Oracle too. Bollocks. There ain't anything special in XPS that isn't also provided (albeit in slightly different ways) by Oracle, DB2 or Teradata, on clusters of MPP technology. XPS, DB2 and Teradata are all based on shared nothing clustering, which requires one approach, Oracle uses shared disk, which requires another. All have been around for 1 or more decades. You pays you money and you takes you choice. -- Regards, Mark Townsend