dual-core opterons

[JPC-Greg]

We have been testing and planned on deploying quad opteron systems shortly. It has come to my attention now that AMD is about to release dual-core processers this month. These processers are compatible with existing 940 socket servers. This is wonderful news but what does it mean?

As if quad opteron machines werent mind boggling enough in speed imagine if you will an 8x opteron system now since we can drop in these new dual-core machines. It would seem instead of just one major fantastci update to quad opteron machines (was going to be a big update and suprise by itself), soon you will be enjoying the bliss of several 8x opteron systems.

We need time to get our ducks in a row on this one but you can expect to see many announcements about servers updating left and right in approx 45-60 days. We have gathered a great deal of data on machines that could use these updates first and we will start there. While all machines are running smooth there are some dual p3 machines in our network that I for one would love to give a jolt of power to. I will be conducting surveys and talking to clients more before making any final upgrades to machines. I first want to see how everyone feels about it. Obviously this would be done with zero downtime and the only thing you the end user sees is a boost in speed.

sample config:
quad dual-core opteron
8gig registered ecc ddr3200
raid 0+1 10k rpm drives

[Vin DSL]

That's great, Chief, but speaking for myself, I'm tethered to a 1.5MB DSL connection, via 2 miles of 'doorbell' wiring to the DSLAM. Sure, I have a 3.0GHz Intel P4 machine sitting on my desktop, but it's all kinda useless 'cause the weakest link in the network is the link between Texas and Arizona.

Raw processing power aside, what is the benefit of this new technology?

[mattsiegman]

ummmm

drool++;

[Piyer]

Jag, if this means that all of LEAD's services won't crawl to a halt anymore every time everyone's cron jobs fire up, then I'm all for it!

[JPC-Greg]

Recall moments of slow php processing on some machines for example. Lead is a prime example and top on the list of this process if/when we move forward. Aside from raw processing power to push site content faster there is a level of consolidation we could do which doesnt yeild a direct noticable difference to clients but does to our staff, which I suppose in turn could show in form of ticket communication, improved uptime, etc.

The main focus is not to start off new signups under these systems in hopes to bring in more signups but rather to take some of our grandfather machines (2-3+ yrs of age) and breath some new life into them. Those machines do not receive new accounts and should be stable unless there is a hardware problem. We will start by taking a few machines like lead (dual p3 1ghz ie 2ghz total with lower FSB) and combing 3 stable machines into one 8x opteron (quad dual-core ie 16ghz total with much higer fsb ) . Even though in this example we combined machines its simply pure overkill processing wise.

Part of the reason for this is over the years a dual p3 just doesnt pack the punch it did when it was new becuase software and hosting needs grow. Programs, daemons, scripts just tend to get more intense and trying to keep current clients updated with these features on older machines means we keep load balancing them to the point where a machine is barely populated. This should remedy that until 3 laters when I imagine everything will catch up to dual-core opterons.

[JPC-Greg]

This upgrade wont reach all clients, at least not in the near future. Vin, if your machine doesnt catch our eye as a potential in need of this change it wont be included.

[Vin DSL]

Interesting...

[Sara]

Mostly I have no clue about all this tech speak, but from what I do understand (I'm not totally blonde) it sounds like good times are coming (hopefully Copper will be among the lucky ones).

[sycopathman]

that is super nice