Answer to the Roof Riddle
In answering the prior question...
As you know, computers consume a ton of energy - if you don't work in a datacenter, you may not know what I'm talking about. But you know how your laptop warms your lap? Or your PC heats up your den? Multiply that a few thousand times over, and you have a problem faced by most datacenters - power draw and heat dissipation. Map that challenge to every business on earth, and you have a global power crisis as the network is built out. (And talk to some web 2.0 startups, you'll hear many say their second biggest operating expense, after salaries, is electricity - that's why the big search companies are building data centers where power's cheap).
Back to my story... the CIO in my prior posting informed her CEO that in order to support more analytics and trading activity - the computational heart of their business - they needed to build a bigger computing grid. For which they needed more space (which isn't cheap in midtown Manhattan), and more power - to which he responded, "the CEO of the power company is a friend of mine - let me just give him a call."
The CIO replied, "no no, those are only a couple of limiters. The more power we bring in, the more cooling we need. The more cooling we need, the more power again. But the thing that's really holding us back is even with more budget for space, power and cooling, we need backup power in the event of an outage, and the generator necessary to provide backup power of this magnitude is the size of a locomotive, and the only place we could possibly put that is on the roof, and look, we DON'T HAVE ANY MORE ROOM!"
And now you know why we've been so focused on the physical size and energy efficiency of our new computing and storage platforms, in addition to their raw performance. (And if you'd like a sample to try for yourself, just click here, or read what others have experienced.) The lowest end systems start at $795 (no, that's not a typo).
_______________
And Mr. Scoble - thanks for taking the time to stop in, much appreciated. I enjoyed the discussion, too. Thank you for pioneering the medium.
Posted on 09:50PM Jun 10, 2006 | Comments[18]



















Posted by Web on June 10, 2006 at 10:39 PM PDT #
Posted by David lewis on June 10, 2006 at 11:23 PM PDT #
Posted by David Carlton on June 11, 2006 at 09:12 PM PDT #
Sun, despite being the first, is not the only company being concerned with the issue. AMD server processors are already reaching the 90 watts range, closing the gap to Sun's T2000.
My question to Jonathan is, do you think the first mover advantage that Sun has will be enough to protect its competitive position in processors and servers that consume less power and produce less heat?
Posted by Future Tech Web on June 12, 2006 at 12:15 AM PDT #
Posted by 194.217.90.28 on June 12, 2006 at 02:17 AM PDT #
Posted by Adam S on June 12, 2006 at 04:38 AM PDT #
Posted by Adriano on June 12, 2006 at 06:33 AM PDT #
Posted by Paul Fiser on June 12, 2006 at 11:05 AM PDT #
So, please, for the love of all that is holy...do me one favor:
Fix your sales team!
When surveying the landscape of our vendor relationships...the one we have with Sun is by far the most broken. We rarely get timely callbacks from our salespeople and they seemingly refuse to discuss the Niagara or Galaxy servers when they do decide to stop in. They seem to fear selling the smaller boxes will eat into their ability to meet their astronomical sales goals.
If you're a salesman who has two kids and a mortgage payment, are you going to try and sell as many high priced, high margin boxes to your customers as you can or are you going to triple the number of required sales calls in order to sell the smaller volume boxes?
For example, I've currently got a web server that is running on an E220r. I'd like to put it on a T1000. Your salesman wants to sell me a V490 and tells me I'm crazy for looking at the T1000, I need more "robust computing". Thanks, but I know what I need. Listen more, talk less bub.
I don't blame him because it makes his life easier but, is that goal of Sun? To make the lives of their salesman easier?
Isn't there a way to simplify the sales goals so they are focused on number of units sold in particular product lines rather than a total dollar value?
Joe Salesman currently has four accounts and a $10MM goal.
Why not make it so Joe Salesman has four accounts and a goal of 100 T1000/T2000 units, 500 X4100/4200 units, and 100 USIV+ units?
Posted by Concerned Customer on June 12, 2006 at 04:03 PM PDT #
Posted by Prince on June 12, 2006 at 04:57 PM PDT #
Posted by Kevin Hutchinson on June 13, 2006 at 12:04 AM PDT #
Posted by Pikemann Urge on June 13, 2006 at 03:29 AM PDT #
Why build a bigger computing grid in Manhattan? Follow the lead of the big search companies, and build a data center where power is cheap.
The space and power savings argument for coolthreads is still a valid and compelling one, and would certainly apply to the new out-of-town datacenter, but it is a separate argument from the "no space on the roof" gimmick that serves as the punchline of this pitch.
Posted by Kevin on June 14, 2006 at 09:46 AM PDT #
Posted by Prince on June 14, 2006 at 03:12 PM PDT #
Posted by Deepak Sekar on June 14, 2006 at 03:40 PM PDT #
Posted by Nathan Bailey on June 15, 2006 at 03:22 AM PDT #
Posted by Peter Denyer on June 15, 2006 at 01:24 PM PDT #
1. The Niagara chips do sound good. However, as one earlier poster pointed out, without Redmond porting Windows to this hardware, you are missing out on a very large opportunity. Can you please work on convincing Redmond to do something about it?
2. Is Solaris ever going to be made user friendly? I recently downloaded Solaris 10 for x86. It installed after a lot of struggle - and the user interface was quite poor. I wonder when we will see results from the Looking Glass project. I have now reformatted and installed Suse with KDE instead. The Java Desktop system is terrible compared to KDE. (I know there is KDE for Solaris, but I wish you bundled it all with an intuitive installer). Learn a few things from Redmond.
3. I use Java everyday, and I think it is a great language. But why can't Sun make money? IBM and BEA (and even Oracle) have milked Java. Don't open source it - but find a way to make money off it.
4. Create a Sun search engine. I am sure Sun's engineers can give Google a good run for its money. Sun could leverage the Java.com site for this purpose. This would be a great showcase for your hardware and software products as well. And finally, it could really add a few dollars to Sun's bottomline.
Posted by Sun Investor on June 15, 2006 at 07:52 PM PDT #