Comments on: ISR G2 – what I wish it was… http://www.douglasgourlay.com/blog/2009/10/isr-g2-what-i-wish-it-was/ Data Centers, Virtualization, and Cloud Computing Sun, 02 May 2010 01:04:25 +0000 hourly 1 http://wordpress.org/?v=3.0 By: Mark http://www.douglasgourlay.com/blog/2009/10/isr-g2-what-i-wish-it-was/comment-page-1/#comment-240 Mark Wed, 28 Oct 2009 22:23:41 +0000 http://www.douglasgourlay.com/blog/?p=146#comment-240 This concept of a virtualized services router (VSR?) popped into my head a couple of years ago. There are issues with the concept (too many network interrupts on a single host OS/hypervisor and processor degrading network performance), but there exist methods and software today which could overcome those issues. Two ideas Sun Microsystems proposed, "Throughput Networking" where a single core of a multicore processor is used as a packet processor for all applications on a box, and "Project Crossbow", which allows a single host OS to support multiple TCP/IP stacks on one kernel, which again eliminates a network bottleneck in a virtualized environment, would make such a platform viable. Throughput networking would work on a multicore x86 processor, and a multiple TCP/IP stack kernel could be adapted to Linux or *BSD OSs. This concept of a virtualized services router (VSR?) popped into my head a couple of years ago. There are issues with the concept (too many network interrupts on a single host OS/hypervisor and processor degrading network performance), but there exist methods and software today which could overcome those issues. Two ideas Sun Microsystems proposed, “Throughput Networking” where a single core of a multicore processor is used as a packet processor for all applications on a box, and “Project Crossbow”, which allows a single host OS to support multiple TCP/IP stacks on one kernel, which again eliminates a network bottleneck in a virtualized environment, would make such a platform viable. Throughput networking would work on a multicore x86 processor, and a multiple TCP/IP stack kernel could be adapted to Linux or *BSD OSs.

]]> By: Douglas Gourlay http://www.douglasgourlay.com/blog/2009/10/isr-g2-what-i-wish-it-was/comment-page-1/#comment-233 Douglas Gourlay Sat, 24 Oct 2009 21:41:15 +0000 http://www.douglasgourlay.com/blog/?p=146#comment-233 Adrian, I disagree a bit. An Intel Nehalem with a decent bit of RAM may actually be a bit more expensive than the current cost-optimized ISR platforms. On any product with that much volume you spend a lot of time and energy ensuring component selection keeps COGS low and that volumes can bring further cost reductions. But the real shifts it would enable are: - decouple memory density from price and get rid of perception of price-gouging while enabling core-value to be delivered through VM based plugins. - another delivery platform for the Nexus 1000v technology - quick time-to-market and time-to-integration of any acquisition with a VM-based offering Thus net-net would be a higher performance, and as technologies integrate, high margin offer with a more extensible hardware platform. That is, if you really wanted the network to be a platform. dg Adrian, I disagree a bit. An Intel Nehalem with a decent bit of RAM may actually be a bit more expensive than the current cost-optimized ISR platforms. On any product with that much volume you spend a lot of time and energy ensuring component selection keeps COGS low and that volumes can bring further cost reductions.

But the real shifts it would enable are:
– decouple memory density from price and get rid of perception of price-gouging while enabling core-value to be delivered through VM based plugins.
– another delivery platform for the Nexus 1000v technology
– quick time-to-market and time-to-integration of any acquisition with a VM-based offering

Thus net-net would be a higher performance, and as technologies integrate, high margin offer with a more extensible hardware platform. That is, if you really wanted the network to be a platform.

dg

]]> By: Adrian M http://www.douglasgourlay.com/blog/2009/10/isr-g2-what-i-wish-it-was/comment-page-1/#comment-232 Adrian M Fri, 23 Oct 2009 18:39:20 +0000 http://www.douglasgourlay.com/blog/?p=146#comment-232 The question is "why should they do it" ? They would lose money. Cheaper boxes would perform better then current ones. The question is “why should they do it” ? They would lose money. Cheaper boxes would perform better then current ones.

]]> By: Carrie http://www.douglasgourlay.com/blog/2009/10/isr-g2-what-i-wish-it-was/comment-page-1/#comment-230 Carrie Fri, 23 Oct 2009 00:42:26 +0000 http://www.douglasgourlay.com/blog/?p=146#comment-230 I hear ya loud and clear! Apt metaphor at the end there. I hear ya loud and clear!
Apt metaphor at the end there.

]]>