Обсуждение: hardware question
Hello *, I've got a question concerning hardware. We're planning to replace our SUNS (E3500,E3000,E450 etc) with multiprocessor XEON boards. Is there any recommendation or experience about this type of hardware ? kind regards -andreas -- Andreas Schmitz - Phone +49 201 8501 318 Cityweb-Technik-Service-Gesellschaft mbH Friedrichstr. 12 - Fax +49 201 8501 104 45128 Essen - email a.schmitz@cityweb.de
There are few items I was wondering if I could get an answer as to whether they are features or bugs and if they are bugs the release in which they might be rectified. Issue #1 If you drop a table and then recreate it. Trying to access a view on that table returns error about OID not being found. If you drop and recreate the view the problem goes away. Issue #2 This one is a bit more indepth but on updates it appears that implicit numeric conversion is not happening. Here is a link to some of our lengthy discussions on it... http://kt.zork.net/GNUe/gnue20020615_33.html#12 I can give exact query examples if the developers dont understand the issue. -- Derek Neighbors GNU Enterprise http://www.gnuenterprise.org derek@gnue.org Was I helpful? Let others know: http://svcs.affero.net/rm.php?r=dneighbo
Вложения
On Thu, Sep 26, 2002 at 08:54:12AM -0700, Derek Neighbors wrote:
>
> If you drop a table and then recreate it. Trying to access a view on
> that table returns error about OID not being found. If you drop and
> recreate the view the problem goes away.
Yes. This is a known problem. I _think_ 7.3 (currently in beta) is
going to do something about it, but don't hold me to that.
>
> Issue #2
>
> This one is a bit more indepth but on updates it appears that implicit
> numeric conversion is not happening. Here is a link to some of our
> lengthy discussions on it...
There has been a great whack of discussion in -hackers recently about
the implicit rules. The short answer is Here There Be Dragons. But
you can probably try quoting everything; you may find it works fine
for you.
A
----
Andrew Sullivan 204-4141 Yonge Street
Liberty RMS Toronto, Ontario Canada
<andrew@libertyrms.info> M2P 2A8
+1 416 646 3304 x110
I assume you'll be running Linux on this? We're running a 4x700 right now, it's very dependable. But be careful, you're getting yourself into (as far as Linux goes) very high end hardware. It's easy enough to find people who can support a uni or dual-proc box. Finding expertise on 4-way or higher systems is more difficult. A wide variety of kernel issues can present themselves. Memory addressing comes to mind. Not many Linux systems have 4+gb of ram, how did you set up bigmem? These things you may not even notice for a year until you grow to tbe point where you hit the choaker. But don't mind me, I've just run into one this last week and it's driving me nuts. (Some weird kernel bug, that is) Of course, none of this matters if you're not running Linux (go you) but then again, I've heard Slowaris x86 is.. well. slow. And you wouldn't be seriously considering NT would you? Does *BSD have those SMP bugs worked out? I would highly recommend you purchase your hardware through a solid company. Get Dell, IBM or Compaq to build you a machine. Xeon machines are very complex, you'll save yourself time, headaches and (lots) of money by going with the big guys. I am curious, why are you getting away from Sun? I've been under the impression that Sun hardware kills IA32 when it comes time to play. (eg. high load) Am I reading too much FUD these days? > -----Original Message----- > From: pgsql-admin-owner@postgresql.org > [mailto:pgsql-admin-owner@postgresql.org]On Behalf Of Andreas Schmitz > Sent: Thursday, September 26, 2002 6:38 AM > To: pgsql-admin@postgresql.org > Subject: [ADMIN] hardware question > > > > Hello *, > > I've got a question concerning hardware. We're planning to > replace our SUNS > (E3500,E3000,E450 etc) with multiprocessor XEON boards. Is there any > recommendation or experience about this type of hardware ? > > > kind regards > > -andreas > > -- > Andreas Schmitz - Phone +49 201 8501 318 > Cityweb-Technik-Service-Gesellschaft mbH > Friedrichstr. 12 - Fax +49 201 8501 104 > 45128 Essen - email a.schmitz@cityweb.de > > > ---------------------------(end of broadcast)--------------------------- > TIP 4: Don't 'kill -9' the postmaster
A note on your last point. We have a bunch of older SUN boxes around here. e450, e250, etc. The 2MB L2 cache un the UltraSparc 2 does allow for a high load. And for their era they were sweet chips, still are. But the #1 reason in my mind to move from Sun to an Lintel box would be cost. Sun is expensive. It seems to me that the Intel/AMD world has moved forward at a faster pace, definelty more bang for the buck (Althons are great). My 2 cents. On Thu, 2002-09-26 at 10:25, Adam Erickson wrote: > I assume you'll be running Linux on this? > > We're running a 4x700 right now, it's very dependable. But be careful, > you're getting yourself into (as far as Linux goes) very high end hardware. > It's easy enough to find people who can support a uni or dual-proc box. > Finding expertise on 4-way or higher systems is more difficult. A wide > variety of kernel issues can present themselves. Memory addressing comes to > mind. Not many Linux systems have 4+gb of ram, how did you set up bigmem? > These things you may not even notice for a year until you grow to tbe point > where you hit the choaker. > > But don't mind me, I've just run into one this last week and it's driving me > nuts. (Some weird kernel bug, that is) Of course, none of this matters if > you're not running Linux (go you) but then again, I've heard Slowaris x86 > is.. well. slow. And you wouldn't be seriously considering NT would you? > Does *BSD have those SMP bugs worked out? > > I would highly recommend you purchase your hardware through a solid company. > Get Dell, IBM or Compaq to build you a machine. Xeon machines are very > complex, you'll save yourself time, headaches and (lots) of money by going > with the big guys. > > I am curious, why are you getting away from Sun? I've been under the > impression that Sun hardware kills IA32 when it comes time to play. (eg. > high load) Am I reading too much FUD these days? > > > -----Original Message----- > > From: pgsql-admin-owner@postgresql.org > > [mailto:pgsql-admin-owner@postgresql.org]On Behalf Of Andreas Schmitz > > Sent: Thursday, September 26, 2002 6:38 AM > > To: pgsql-admin@postgresql.org > > Subject: [ADMIN] hardware question > > > > > > > > Hello *, > > > > I've got a question concerning hardware. We're planning to > > replace our SUNS > > (E3500,E3000,E450 etc) with multiprocessor XEON boards. Is there any > > recommendation or experience about this type of hardware ? > > > > > > kind regards > > > > -andreas > > > > -- > > Andreas Schmitz - Phone +49 201 8501 318 > > Cityweb-Technik-Service-Gesellschaft mbH > > Friedrichstr. 12 - Fax +49 201 8501 104 > > 45128 Essen - email a.schmitz@cityweb.de > > > > > > ---------------------------(end of broadcast)--------------------------- > > TIP 4: Don't 'kill -9' the postmaster > > > ---------------------------(end of broadcast)--------------------------- > TIP 6: Have you searched our list archives? > > http://archives.postgresql.org
On Thu, Sep 26, 2002 at 12:25:51PM -0500, Adam Erickson wrote:
> I am curious, why are you getting away from Sun? I've been under the
> impression that Sun hardware kills IA32 when it comes time to play. (eg.
> high load) Am I reading too much FUD these days?
Maybe. I can say for sure that an older PIII 2-way VA Linux box
running FreeBSD that we had here was running cirles around a 2-way
E450 we have. I think it sort of depends on the kind of load.
The really _big_ advantage Sun is supposed to offer is the supreme
hardware reliability. In the interests of not rousing any lawyers, I
will refrain from commenting on the extent to which that advantage is
realised in fact.
A
--
----
Andrew Sullivan 204-4141 Yonge Street
Liberty RMS Toronto, Ontario Canada
<andrew@libertyrms.info> M2P 2A8
+1 416 646 3304 x110
> I've got a question concerning hardware. We're planning to replace our SUNS > (E3500,E3000,E450 etc) with multiprocessor XEON boards. Is there any > recommendation or experience about this type of hardware ? A lot of it depends on whether you're talking about P3 Xeons or P4 Xeons. P3 Xeons are not worth the cost, in my not-so-humble opinion. We replaced a 4x700 MHz/1MB P3 Xeon machine with a 2x AthlonMP 2000+, and performance increased substantially, dropping the system loads in *half*. If you're talking P4 Xeons, I can't comment on their effectiveness vs. AthlonMP's - but they sure do come at a price premium. steve
Hello *, thanks a lot for the quick responses. It looks like I have to answer a few questions. - I am talking about P4 Xeon - Getting away from SUN is the attemp to reduce costs. I think I don't have to mention the amount of money that has to be spent on service contracts (even with older machines). I don't really like to change from multi CPU SUNs to Intel boxes. My Problem is that I don't know much about the behavior of Intel Xeons and I fear to run into a performance bottleneck even with databases >5Gb. I've one P4 Dual Xeon box running with oracle 9i. The system shows 4 CPUs that seems to be typical for Xeon. Will postgres use this a a 4 CPU machine ? Is there any HT support ? regards -andreas -- Andreas Schmitz - Phone +49 201 8501 318 Cityweb-Technik-Service-Gesellschaft mbH Friedrichstr. 12 - Fax +49 201 8501 104 45128 Essen - email a.schmitz@cityweb.de
Here's a bit more anecdotal info on multi-processor systems- We've got one system with Tyan motherboard & 2 1.2 GHz Athlon processors, SCSI Raid, 1 GB RAM. We're running Debian Linux 2.4.17 & Postgres 7.2.1. We're happy with the performance, & both processors are being used by postgresql. We haven't seen all of the RAM used yet, but I'm confident that the OS is equipped to use it & just think we haven't challenged it enough. On the downside, our backup machine with the same configuration doesn't work. Since the other one is performing flawlessly, I'm assuming we have a bad hardware component somewhere. I've also run a 4 processor Xeon box with 2GB RAM at a previous job, and I wasn't impressed, but it was running NT & Oracle so the comparison is very approximate. I would encourage you to strongly consider the Athlon processors if you're trying to reduce cost. From what I've read elsewhere, it sounds like you may get comparable performance with a dual Athlon vs a Quad Xeon. Also remember that in a server-class Sun box, you've got a nice I/O channel by default, but in an Intel/AMD box, you may need to think hard about I/O & make some choices there as well. -Nick > -----Original Message----- > From: pgsql-admin-owner@postgresql.org > [mailto:pgsql-admin-owner@postgresql.org]On Behalf Of Andreas Schmitz > Sent: Friday, September 27, 2002 3:43 AM > To: pgsql-admin@postgresql.org > Subject: Re: [ADMIN] hardware question > > > > Hello *, > > thanks a lot for the quick responses. It looks like I have to > answer a few > questions. > > - I am talking about P4 Xeon > - Getting away from SUN is the attemp to reduce costs. I think I > don't have to > mention the amount of money that has to be spent on service > contracts (even > with older machines). > > I don't really like to change from multi CPU SUNs to Intel boxes. > My Problem > is that I don't know much about the behavior of Intel Xeons and I > fear to run > into a performance bottleneck even with databases >5Gb. I've one > P4 Dual Xeon > box running with oracle 9i. The system shows 4 CPUs that seems to > be typical > for Xeon. Will postgres use this a a 4 CPU machine ? Is there any > HT support > ? > > regards > > -andreas > > > -- > Andreas Schmitz - Phone +49 201 8501 318 > Cityweb-Technik-Service-Gesellschaft mbH > Friedrichstr. 12 - Fax +49 201 8501 104 > 45128 Essen - email a.schmitz@cityweb.de > > > ---------------------------(end of broadcast)--------------------------- > TIP 4: Don't 'kill -9' the postmaster >