Have your new servers been running long enough to compare the aggregate amount of power your old and new servers consume in an average month? Your old p4 servers should have consolidated many to one into the new servers (or if still 1:1 operate much closer to idle); so I suspect the total savings are much larger than the 2.5-5:1 that benching a single new server vs a single old server shows; the latter number would be even more convincing from a "this is why we should spend money on new servers even though our six year old ones haven't died yet" perspective.
"The difference at idle however is far more impressive however"
I think you have too many "howevers" however if you just really like to say however I can however understand however silly it may sound you just want to really drive home the point:)
You're saying you'll deploy 6 more of the Westmere boxes this year. Is this a good idea? Having similar hardware is surely nice, but there's newer stuff out by now. Depends on what your sponsor is willing to give you, though.
My assumption would be they're setting up a redundant copy of the site in a different data center. Using identical hardware in both locations would make things simpler for that purpose.
we faced much the same quandry last year. the decisions were made to roll our own with much the same performance increases and power reductions as the Anandtech article shows. 1 - we're sort of gear heads and wanted to make this ours 2 - there is a loss of control when you push stuff to the cloud, even to Amazon. 3 - there may be security and data issues which mandate keeping the systems under closer control. 4 - it's fun to do this! this is their business model and how they make the revenue side work. 5 - they can always push data backup and replication of software or hardward and data to Amazon if they want to do that llater.
Great article. Of the three (ssd/cpu/power) i thought for sure i'd like be most impressed by your SSD architecture and the improvements they bring in IO. But after reading all three it turns out i found this to be the most fascinating and thrilling example of technological improvement.
More performance for less energy. I can't wait to see what the future brings.
We’ve updated our terms. By continuing to use the site and/or by logging into your account, you agree to the Site’s updated Terms of Use and Privacy Policy.
10 Comments
Back to Article
DanNeely - Monday, March 18, 2013 - link
Have your new servers been running long enough to compare the aggregate amount of power your old and new servers consume in an average month? Your old p4 servers should have consolidated many to one into the new servers (or if still 1:1 operate much closer to idle); so I suspect the total savings are much larger than the 2.5-5:1 that benching a single new server vs a single old server shows; the latter number would be even more convincing from a "this is why we should spend money on new servers even though our six year old ones haven't died yet" perspective.Hrel - Monday, March 18, 2013 - link
"The difference at idle however is far more impressive however"I think you have too many "howevers" however if you just really like to say however I can however understand however silly it may sound you just want to really drive home the point:)
MrSpadge - Monday, March 18, 2013 - link
You're saying you'll deploy 6 more of the Westmere boxes this year. Is this a good idea? Having similar hardware is surely nice, but there's newer stuff out by now. Depends on what your sponsor is willing to give you, though.marc1000 - Monday, March 18, 2013 - link
they deployed it on 2010..."As I mentioned in my first post, we went down this path back in 2010"
Gigaplex - Monday, March 18, 2013 - link
They're also looking to deploy 6 more this year."with another 6 to be deployed this year"
DanNeely - Tuesday, March 19, 2013 - link
My assumption would be they're setting up a redundant copy of the site in a different data center. Using identical hardware in both locations would make things simpler for that purpose.mayankleoboy1 - Tuesday, March 19, 2013 - link
I would hate to compare the performance/watt difference between these two systems.bwanaaa - Tuesday, March 19, 2013 - link
why did anandtech decide to roll their own and not use amazon's infrastructure? is it cheaper to roll your own?pensive69 - Wednesday, May 15, 2013 - link
we faced much the same quandry last year.the decisions were made to roll our own with much the same
performance increases and power reductions as the Anandtech
article shows.
1 - we're sort of gear heads and wanted to make this ours
2 - there is a loss of control when you push stuff to the cloud, even to Amazon.
3 - there may be security and data issues which mandate keeping the
systems under closer control.
4 - it's fun to do this! this is their business model and how they make
the revenue side work.
5 - they can always push data backup and replication of software or
hardward and data to Amazon if they want to do that llater.
alacard - Saturday, March 23, 2013 - link
Great article. Of the three (ssd/cpu/power) i thought for sure i'd like be most impressed by your SSD architecture and the improvements they bring in IO. But after reading all three it turns out i found this to be the most fascinating and thrilling example of technological improvement.More performance for less energy. I can't wait to see what the future brings.