Re: Memory Management

On 02-09-2010 20:24, shorti wrote: > On Sep 2, 12:06 pm, Andrew Ford<af...@networkip.net> wrote: >>> ok..then I will give you another chance to redeem yourself.....Can >>> anyone tell me what they think of Informix Performance for a larger >>> database...lets say 5-6 + million rows on the main table with multiple >>> other tables with the same amount of rows. Is it better to go with >>> something like Oracle or DB2 instead. I know, I know...I am asking in >>> the Informix forum but seriously..does Informix measure up when it >>> comes to larger databases? >> >> My largest OLTP table contains 282 million rows and the engine handles about >> 3 million transactions per day using 2 dual core CPUs and 3GB of memory for >> Informix with an average CPU idle of 80-90%. I would classify this as a >> small to medium sized Informix install, there are much larger Informix >> installs out there. 5-6 million rows should not be a problem. >> >> Andrew > > Thanks all..great information. I agree that 6 Million is not really > that big since the big RDBMSs can navigate 100's of millions of > records but, sorry to say, Informix isnt on the list of the top RDBMSs > so I wasnt sure. I will have a different view of Informix from now > on. I do have to say that I am pretty impressed by all that I have > researched so far and puzzled why it is not in the running with > Oracle, DB2 and MS. > Because in the current "fast" world nobody does what you're doing... I'd say people do this associations: MS -> Cheap & easy Oracle -> Leader DB2 -> IBM (for better and for worst) Congratulations for your achievement... Regards.

> From: domusonline@gmail.com > Subject: Re: Memory Management > Date: Sun, 5 Sep 2010 23:32:32 +0100 > To: informix-list@iiug.org > > > Thanks all..great information. I agree that 6 Million is not really > > that big since the big RDBMSs can navigate 100's of millions of > > records but, sorry to say, Informix isnt on the list of the top RDBMSs > > so I wasnt sure. I will have a different view of Informix from now > > on. I do have to say that I am pretty impressed by all that I have > > researched so far and puzzled why it is not in the running with > > Oracle, DB2 and MS. > > > > Because in the current "fast" world nobody does what you're doing... > I'd say people do this associations: > > MS -> Cheap & easy > Oracle -> Leader > DB2 -> IBM (for better and for worst) > Sigh. Define by what you mean "fast". If you're talking strictly RDBMS for OLTP (read TPC-C type benchmarking), Informix probably blows the competition away on the Intel e5500+ series. The reason I say probably is that well, IBM doesn't *do* benchmarks using Informix. (Even though Jerry K. hinted at it some time ago but was again forced to back peddle because well IDS isn't DB2.) (And nobody but mainframe dominated shops want DB2 LUW.) Imagine this... 2 socket 4 core e5500 series Intel chips. (This is *now* a commodity hardware platform BTW...) Put this in to a 4U high box, and two "high end" SAS/SATA controller that can do 6.0 gbs (SATA III). I forget who posted this out of Germany but you're better off using SSDs over the FusionIO cards, so taking his word, go with 2.5" SSDs. (I believe that SATA III drive are actually cheaper than SATA II now...) If the motherboard doesn't have 10GBe, you can get a seriously good card from Solarflare. Fully loaded a machine would set you back $12-15K fully loaded with 64-96GB of memory. Go run your TPC-C and calculate your cost per transaction. Your most expensive component will be the cost of the IDS engine running on 8 cores. IDS will be able to take advatage of the multiple cores better than any other RDBMs on the market. But if by fast you mean DW and processing through hundreds of TB of information, or even a PB of information, then no. IDS is not "fast". You'll need to go to a distributed system which means NOSQL Hadoop/HBase depending on what you're trying to do. And with Pig now able to hit against HBase... well. Yes, you can get fast. Google/Yahoo!/Facebook/Twitter/etc ... fast. MS isn't really cheap or easy when you consider that it doesn't scale and on the low end you could go MySQL or Postgres for a lower TCO. (You have to pay for Windows 7 ;-) But getting back to the RDBMS world... We don't know how fast because IDS isn't bench marked. There's more to this, but why talk about something where it doesn't matter. BTW if I were to go IDS, I'd build an Apple OS/X server or set of them so I could run the full enterprise release for free. (Assuming that IBM is still updating and maintaining the Apple release. ) ;-) (The downside is that the hardware is more expensive, but when you consider the savings on Software costs... still cheaper. ;-) -G

Hello Fernando, I hate to disappoint you, however MS is not easy; try to restore a complete level 0 backup, you will see what a can of wurms that is. Besides that monitoring and trying to grab the hottest sql is not easy(sqlsrv2005); the stuff described can be used; however when a query ends, then the needed info is stored in the relevant tables so you will never see the hottest sql in action that also may cause you to miss a problem sql. Obstacle is a leader; yeah however i still can not understand why people accept the pile of shit. online backup one needs the trx log and createing / restoreing looks like how informix used to do this with onarchive which was the worst piece of shit ever produced by informix. The data you gonna get in your backup is when the backup ends. Missing set lock mode to wait xxx DB2CV well i have not touched this nor come across this for a long time fortunatly, if you add a container does it still redistribute all data in your tablespace??? and the equiv of xps, does it still give you an error when you update a row which should cause the row to move to another partition; if you unload a table with blobs are blobs still written in file.XXX where XXX goes from 000 to 999. set lock mode to wait for the whole instance?? Informix is way fast and easy!!!!!! Dono about priceing; am not a bean counter. Superboer.

On Tue, Sep 7, 2010 at 8:27 AM, Superboer <superboer7@t-online.de> wrote: > Hello Fernando, > > I hate to disappoint you, however MS is not easy; try to restore a > complete level 0 backup, you will see what a can of wurms that is. > Besides that monitoring and trying to grab the hottest sql is not > easy(sqlsrv2005); the stuff described can be used; however when a > query ends, > then the needed info is stored in the relevant tables so you will > never see the hottest sql in action that also may cause you to miss a > problem sql. > > Obstacle is a leader; yeah however i still can not understand why > people accept the pile of shit. > online backup one needs the trx log and createing / restoreing looks > like how informix used to do this with onarchive which was the worst > piece of shit ever produced by informix. The data you gonna get in > your backup is when the backup ends. > Missing set lock mode to wait xxx > > DB2CV well i have not touched this nor come across this for a long > time fortunatly, if you add a container does it still redistribute all > data > in your tablespace??? and the equiv of xps, does it still give you an > error when you update a row which should cause the row to move to > another > partition; if you unload a table with blobs are blobs still written in > file.XXX where XXX goes from 000 to 999. > set lock mode to wait for the whole instance??> > > Informix is way fast and easy!!!!!! Dono about priceing; am not a bean > counter. > > > Superboer. > > _______________________________________________ > Informix-list mailing list > Informix-list@iiug.org > http://www.iiug.org/mailman/listinfo/informix-list > You did not disappoint me :) What I wrote is not my opinion. It's what I think the market thinks. And as usual (contrary to other posters :) ) I may be wrong! But it's always nice to see others opinions, specially when they are friendly to Informix :) Regards. -- Fernando Nunes Portugal http://informix-technology.blogspot.com My email works... but I don't check it frequently...