> For my application there is very little info I can share. Maybe less than 10 on 100 actually so I not sure it worth it ... Ok, so 90% of the tables are being written to - this either means that your application uses very little constants, or that it has access to constans that are stored somewhere else (eg, a JMX Mbean that's initialized from property files on application startup). Would it be too much work to redesign the DB model to support more than one client? >I look into the HP DL385 and DL585 on HP site and they are price between >3000 and 15000$$ (base price). Thats quite a difference? So is the HP >DL385 with 2 cpus will do the job ? Yeah, there's quite a difference on the price tags between those two. I'd vote for the DL385 since the sockets for the two extra CPU's won't give you linear scalability per $ in the end. A single machine may be cheaper to administrate, but if administration costs are irrelevant/negligible I'd go for several 2-socket machines instead of one 4-socket machine. /Mikael > -----Original Message----- > From: pgsql-performance-owner@xxxxxxxxxxxxxx > [mailto:pgsql-performance-owner@xxxxxxxxxxxxxx] On Behalf Of David > Gagnon > Sent: den 3 juli 2006 13:42 > To: pgsql-performance@xxxxxxxxxxxxxx > Subject: [PERFORM] Is postgresql ca do the job for software deployed in > ASP ou SaaS mode? > > Hi all, > > I've been working on my personal project for 3.5 years now. I > developed an ERP system in web/java. Now the people I will work with > suggest to offers it in Saas mode. Which means my customer will connect > to my website and found they ERP software and data there. It's not the > deployment I planned initially so if you can just validate some > technicals points to be sure it's not crazy using Postgresl here and not > a big $$$ db to do the job. > > Typically I will have 1db per client and around 150 tables per db. So > since I hope I didn`t work all those year for nothing .. I expect to > have bunch of clients witch means the same amount of db since I have 1 > db/client. > > Can I hope having several hundred of db on 1 db server? Like 250 dbs = > 250 client = 360 000 tables !!! > So is there a limit for the number of db in the db server ?(this spec is > not on the website) What about the performance? Can I expect to have the > same performance? > > Since I put everything on the web I do needs an High Availability > infrastructure. I looked into SlonyI and Mammoth to replicate the db > but since SlonyI use triggers can I expect it to do the job? Is Mammoth > is the only available solution? > > Last question and not the least I'm reading this performance list for > several years now and know suggestion about hardware to run postgresl is > discussed. Since I wrote software there severals points about hardware > that I don`t understand. Do you have any suggestion of platform to run > into my Saas configuration? I do need the WISE one! I'm pretty sure > that if I was a big company I would be able throw bunch of $$$$ but it's > not my case. I'm pretty sure it exists out there some piece of Hardware > that would do the job perfectly with a fair price. > > So far I did understand that Postgresql loves Opteron and I have looked > into the dl145 series of HP. I did understand that Dell Hardware it`s > not reliable. But it's still not clear what should be my requirement > for memory, disk, nb cpu, cpu power, etc. > > I'm pretty sure it`s better to have more slower CPUs that having the > latest Opteron available on the market, or more slower servers that > having the fastest one... am I right? But agains what it`s the optimal > choice? > > Thanks you to share your knowledge on those point. I do consider using > Postgresql is the Smart choice in my project since the beginning but > before putting all the money (That I don`t have ..:-)) to buy some > hardware I just want to be sure I'm not crazy! > > Thanks for your help I really appreciate it!! > > Best Regards > /David > > > > > > > > ---------------------------(end of broadcast)--------------------------- > TIP 4: Have you searched our list archives? > > http://archives.postgresql.org > > > > >
