Re: Reliability recommendations

At 11:21 AM 2/15/2006, Jeremy Haile wrote:
>We are a small company looking to put together the most cost effective
>solution for our production database environment. Currently in
>production Postgres 8.1 is running on this machine:
>
>Dell 2850
>2 x 3.0 Ghz Xeon 800Mhz FSB 2MB Cache
>4 GB DDR2 400 Mhz
>2 x 73 GB 10K SCSI RAID 1 (for xlog and OS)
>4 x 146 GB 10K SCSI RAID 10 (for postgres data)
>Perc4ei controller
>
>The above is a standard Dell box with nothing added or modified beyond
>the options available directly through Dell. We had a bad processor last
>week that effectively put us down for an entire weekend. Though it was
>the web server that failed, the experience has caused us to step back
>and spend time coming up with a more reliable/fail-safe solution that
>can reduce downtime.
>
>Our load won't be substantial so extreme performance and load balancing
>are not huge concerns. We are looking for good performance, at a good
>price, configured in the most redundant, high availability manner
>possible. Availability is the biggest priority.
>
>I sent our scenario to our sales team at Dell and they came back with
>all manner of SAN, DAS, and configuration costing as much as $50k.
>
>We have the budget to purchase 2-3 additional machines along the lines
>of the one listed above. As a startup with a limited budget, what would
>this list suggest as options for clustering/replication or setting our
>database up well in general?

1= Tell Dell "Thanks but no thanks." and do not buy any more
equipment from them. Their value per $$ is less than other options
available to you.

2= The current best bang for the buck HW (and in many cases, best
performing as well) for pg:
a= AMD K8 and K9 (dual core) CPUs. Particularly the A64 X2 3800+
when getting the most for your $$ matters a lot
pg gets a nice performance boost from running in 64b.
b= Decent Kx server boards are available from Gigabyte, IWill,
MSI, Supermicro, and Tyan to name a few.
IWill has a 2P 16 DIMM slot board that is particularly nice
for a server that needs lots of RAM.
c= Don't bother with SCSI or FC HD's unless you are doing the most
demanding kind of OLTP. SATA II HD's provide better value.
d= HW RAID controllers are only worth it in certain
scenarios. Using RAID 5 almost always means you should use a HW RAID
controller.
e= The only HW RAID controllers worth the $$ for you are 3ware
Escalade 9550SX's and Areca ARC-11xx or ARC-12xx's.
*For the vast majority of throughput situations, the ARC-1xxx's
with >= 1GB of battery backed WB cache are the best value*
f= 1GB RAM sticks are cheap enough and provide enough value that
you should max out any system you get with them.
g= for +high+ speed fail over, Chelsio and others are now making
PCI-X and PCI-E 10GbE NICs at reasonable prices.
The above should serve as a good "pick list" for the components of
any servers you need.

3= The most economically sound HW and SW architecture that best suits
your performance and reliability needs is context dependent to your
specific circumstances.

Where are you located?
Ron