for clustered server architecture, check out TurboLinux Cluster Server (http://beta.turbolinux.com/cluster/). it's in its last beta stages now. we're learning both postgres and turbocluster; postgres can't handle the distributed architecture right now (locks in shared memory = single-cpu), but hopefully soon? bayard pgsql-admin-digest wrote: > pgsql-admin-digest Thursday, October 28 1999 Volume 01 : Number 373 > > Index: > > HSA (Highly Scalable Architecture) Distribution and replication > Re: [ADMIN] HSA (Highly Scalable Architecture) Distribution and replication > To many clients > Re: [ADMIN] To many clients > Re: [ADMIN] To many clients > New proceses > limit to tables in database. > Re: [ADMIN] HSA (Highly Scalable Architecture) Distribution and replication > > ---------------------------------------------------------------------- > > Date: Thu, 28 Oct 1999 01:11:32 -0700 > From: Rich Ryan <[EMAIL PROTECTED]> > Subject: HSA (Highly Scalable Architecture) Distribution and replication > > Has anyone written a HSA e-commerce solution using postgres? In english, > this means I want more than one server running postgres with the same > database(s). I get lots and lots of queries, and it's just too much for one > poor machine to handle. I get far less inserts, updates, and deletions, but > regardless, the DB's have to be synchronized every couple hours. If I can't > find any HSA to start from, then I'll start from scratch, but I thought I'd > ask first. If I start from scratch, an initial architecture that comes to > mind is designating one server as the master, and the others as slaves. > Every slave gives the rows to be replicated (since the last replication) to > the master at a configurable interval. When the master has been served by > all the slaves, he then synchronizes everyone...I read somewhere that this > is one of the advantages Oracle has over Postgres, but that in a year > Postgres should have an HSA solution. Is anyone working on this? Maybe we > could start? > Thanks, > Rich > > ------------------------------ > > Date: Thu, 28 Oct 1999 09:31:41 +0100 > From: Peter Galbavy <[EMAIL PROTECTED]> > Subject: Re: [ADMIN] HSA (Highly Scalable Architecture) Distribution and replication > > On Thu, Oct 28, 1999 at 01:11:32AM -0700, Rich Ryan wrote: > > Has anyone written a HSA e-commerce solution using postgres? In english, > > this means I want more than one server running postgres with the same > > database(s). I get lots and lots of queries, and it's just too much for one > > poor machine to handle. I get far less inserts, updates, and deletions, but > > regardless, the DB's have to be synchronized every couple hours. If I can't > > find any HSA to start from, then I'll start from scratch, but I thought I'd > > ask first. If I start from scratch, an initial architecture that comes to > > mind is designating one server as the master, and the others as slaves. > > Every slave gives the rows to be replicated (since the last replication) to > > the master at a configurable interval. When the master has been served by > > all the slaves, he then synchronizes everyone...I read somewhere that this > > is one of the advantages Oracle has over Postgres, but that in a year > > Postgres should have an HSA solution. Is anyone working on this? Maybe we > > could start? > > There must be better ways of doing it, but we thought of using a > transaction log from a *single* central update server (writer) to > update multiple readers, running in '-F' mode. Not yet tried it. > > Regards, > - -- > Peter Galbavy > Knowledge Matters Ltd > http://www.knowledge.com/ > > ------------------------------ > > Date: Thu, 28 Oct 1999 10:40:57 +0200 > From: Grzegorz =?iso-8859-1?Q?Prze=9Fdziecki?= <[EMAIL PROTECTED]> > Subject: To many clients > > I have one error > Sorry, too many clients alredy > > What I have to do > Help me I realy need > > Best regards > > ------------------------------ > > Date: Thu, 28 Oct 1999 12:24:27 +0200 (MET DST) > From: Peter Eisentraut <[EMAIL PROTECTED]> > Subject: Re: [ADMIN] To many clients > > Start the postmaster with the -N option and give it a bigger argument. The > default is 32. For example: > > postmaster -N 64 ... > > On Thu, 28 Oct 1999, Grzegorz Prze�dziecki wrote: > > > > > I have one error > > Sorry, too many clients alredy > > > > What I have to do > > Help me I realy need > > > > Best regards > > > > > > ************ > > > > > > - -- > Peter Eisentraut Sernanders vaeg 10:115 > [EMAIL PROTECTED] 75262 Uppsala > http://yi.org/peter-e/ Sweden > > ------------------------------ > > Date: Thu, 28 Oct 1999 12:41:39 +0200 > From: Grzegorz =?iso-8859-2?Q?Prze=BCdziecki?= <[EMAIL PROTECTED]> > Subject: Re: [ADMIN] To many clients > > Change -B too > Thanks a lot > > Peter Eisentraut wrote: > > > Start the postmaster with the -N option and give it a bigger argument. The > > default is 32. For example: > > > > postmaster -N 64 ... > > > > On Thu, 28 Oct 1999, Grzegorz Prze�dziecki wrote: > > > > > > > > I have one error > > > Sorry, too many clients alredy > > > > > > What I have to do > > > Help me I realy need > > > > > > Best regards > > > > > > > > > ************ > > > > > > > > > > -- > > Peter Eisentraut Sernanders vaeg 10:115 > > [EMAIL PROTECTED] 75262 Uppsala > > http://yi.org/peter-e/ Sweden > > > > ************ > > ------------------------------ > > Date: Thu, 28 Oct 1999 16:21:53 +0200 > From: Grzegorz =?iso-8859-1?Q?Prze=9Fdziecki?= <[EMAIL PROTECTED]> > Subject: New proceses > > Welcom > > Few new question > > First > When new user proces on Linux server is created when client is MS > Access. > > Second > I have problem with memory. > Server 64 MbRAM :|) and 150Mb Swap memory. > When start about 30th postgres proceses the postmaster can't alocate > memory and finish work. It is normal ? > Alocate about 5Mb Swap memory only. > > I know SQL Language and what's database but other thema are not clear > for me. > > Best Regards > > ------------------------------ > > Date: Thu, 28 Oct 1999 11:48:34 -0400 > From: "Eric Naujock " <[EMAIL PROTECTED]> > Subject: limit to tables in database. > > Is there a limit to the number of tables that can exist in a particuar database? > > If so what is that limit? or is it largely a function of memeory and storage. > > ------------------------------ > > Date: Thu, 28 Oct 1999 08:50:34 -0700 > From: Rich Ryan <[EMAIL PROTECTED]> > Subject: Re: [ADMIN] HSA (Highly Scalable Architecture) Distribution and replication > > I think having only one writer greatly simplifies things, but it also > provides a single point of failure, which I think kills the idea for me. > I've had too many experiences with network trouble in different areas where > the west coast goes down, but east coast is still reachable, or Dallas goes > out or something. > > - ----- Original Message ----- > From: Peter Galbavy <[EMAIL PROTECTED]> > To: Rich Ryan <[EMAIL PROTECTED]> > Cc: <[EMAIL PROTECTED]> > Sent: Thursday, October 28, 1999 1:31 AM > Subject: Re: [ADMIN] HSA (Highly Scalable Architecture) Distribution and > replication > > > On Thu, Oct 28, 1999 at 01:11:32AM -0700, Rich Ryan wrote: > > > Has anyone written a HSA e-commerce solution using postgres? In english, > > > this means I want more than one server running postgres with the same > > > database(s). I get lots and lots of queries, and it's just too much for > one > > > poor machine to handle. I get far less inserts, updates, and deletions, > but > > > regardless, the DB's have to be synchronized every couple hours. If I > can't > > > find any HSA to start from, then I'll start from scratch, but I thought > I'd > > > ask first. If I start from scratch, an initial architecture that comes > to > > > mind is designating one server as the master, and the others as slaves. > > > Every slave gives the rows to be replicated (since the last replication) > to > > > the master at a configurable interval. When the master has been served > by > > > all the slaves, he then synchronizes everyone...I read somewhere that > this > > > is one of the advantages Oracle has over Postgres, but that in a year > > > Postgres should have an HSA solution. Is anyone working on this? Maybe > we > > > could start? > > > > There must be better ways of doing it, but we thought of using a > > transaction log from a *single* central update server (writer) to > > update multiple readers, running in '-F' mode. Not yet tried it. > > > > Regards, > > -- > > Peter Galbavy > > Knowledge Matters Ltd > > http://www.knowledge.com/ > > > > ************ > > ------------------------------ > > End of pgsql-admin-digest V1 #373 > ********************************* > > ************ ************
