http://storagemojo.com/2006/07/28/architecting-the-internet-data-center-pt-iii/ Data storage info & analysis Tue, 07 Feb 2012 16:02:02 +0000 hourly 1 http://wordpress.org/?v=3.3.1 http://storagemojo.com/2006/07/28/architecting-the-internet-data-center-pt-iii/comment-page-1/#comment-4453 Robin Harris Sat, 29 Jul 2006 22:27:48 +0000 http://storagemojo.com/?p=203#comment-4453 This may be why Google uses 100 Mb ethernet - at least as of three years ago. Wouldn't surprise me if they still were - the 3x bandwidth improvement normally seen over fast ethernet hardly seems worth it. This may be why Google uses 100 Mb ethernet – at least as of three years ago. Wouldn’t surprise me if they still were – the 3x bandwidth improvement normally seen over fast ethernet hardly seems worth it.

]]> http://storagemojo.com/2006/07/28/architecting-the-internet-data-center-pt-iii/comment-page-1/#comment-4442 Robert Pearson Sat, 29 Jul 2006 18:15:55 +0000 http://storagemojo.com/?p=203#comment-4442 My compliments on the "Architecting The Internet Data Center", especially Pt III. Why Pt III? It has been my observation that the state-of-the-art of the "Enabling Technology" is several orders of magnitude ahead of IT thinking. You can always process more Information than you can supply. I have been working on this from the Storage side. Hence my "Speed Limit of the Information Universe" is Storage oriented not Processing oriented. My "storage pyramid" is inverted from the venerable concept in your article and ordered by Content. Lots of important Information waits on less important Information, based on Content. Hence the inversion. It strikes me your IDC model is OLTP focused. That's nice work if you can get it. IMHO, there seems to be a lot more non-OLTP Processing in Ad-Hoc Information spaces. Of course there are all those server pages generated in the IDC? My focus is on the Content of the Information entity or object. To get around the naming conflict I use the term "Managed Unit of Information" for the Content an Information entity or object. What are the demands on this Managed Unit of Information? One demand is Findability Search, Find and Obtain network requests. Quoting from Wikipedia, who say it better than I, "The TCP software implementations on host systems require extensive computing power. Gigabit TCP communication using software processing alone is enough to fully load a 2.4 GHz Pentium 4 processor, resulting in little or no processing resources left for the applications to run on the system. As of 2006, very few consumer network interface cards support TOE." If the processor bandwidth is already exhausted by your OLTP scenario, what do I do for my needs? I prayed to the network Gods for years for TCP/IP rewrites. Then I prayed to the Broadband God and he delivered Gigabit Ethernet. Then I prayed to the InfiniBand God without success. Now I pray to the TOE God along with the Access Density God. There is no TCP/IP God. My compliments on the “Architecting The Internet Data Center”, especially Pt III.
Why Pt III?
It has been my observation that the state-of-the-art of the “Enabling Technology” is several orders of magnitude ahead of IT thinking. You can always process more Information than you can supply.
I have been working on this from the Storage side. Hence my “Speed Limit of the Information Universe” is Storage oriented not Processing oriented. My “storage pyramid” is inverted from the venerable concept in your article and ordered by Content. Lots of important Information waits on less important Information, based on Content. Hence the inversion.
It strikes me your IDC model is OLTP focused. That’s nice work if you can get it. IMHO, there seems to be a lot more non-OLTP Processing in Ad-Hoc Information spaces. Of course there are all those server pages generated in the IDC?
My focus is on the Content of the Information entity or object. To get around the naming conflict I use the term “Managed Unit of Information” for the Content an Information entity or object.
What are the demands on this Managed Unit of Information?
One demand is Findability Search, Find and Obtain network requests.
Quoting from Wikipedia, who say it better than I, “The TCP software implementations on host systems require extensive computing power. Gigabit TCP communication using software processing alone is enough to fully load a 2.4 GHz Pentium 4 processor, resulting in little or no processing resources left for the applications to run on the system. As of 2006, very few consumer network interface cards support TOE.”
If the processor bandwidth is already exhausted by your OLTP scenario, what do I do for my needs?
I prayed to the network Gods for years for TCP/IP rewrites. Then I prayed to the Broadband God and he delivered Gigabit Ethernet. Then I prayed to the InfiniBand God without success. Now I pray to the TOE God along with the Access Density God. There is no TCP/IP God.

]]>