Comments on: The “Power” of a SAN Switch http://blogs.cisco.com/datacenter/the_power_of_a_san_switch/ Wed, 22 May 2013 20:06:20 +0000 hourly 1 By: Dante Malagrinò http://blogs.cisco.com/datacenter/the_power_of_a_san_switch/#comment-10371 Dante Malagrinò Thu, 22 May 2008 19:35:30 +0000 http://blogs.cisco.com/the_fallacy_of_wire_rate_switching_in_the_data_center#comment-10371 I stand corrected by Brian and I apologize for the inaccuracies in the original post. I guess I have known the ESG for so long that I did not notice they had changed the spelling of their name ;-) I also confirm that the ESG did not actually test the configurations, while their report is based on technical review and analysis. The actual testing”" I was referring to has been conducted in our internal labs at Cisco and can be verified by customers during EBC (Executive Briefing Center) visits in San Jose.In the meanwhile, I must say that power and cooling continues to be a top priority for us and for most of the customers I deal with daily.Stay tuned!”

   0 likes

]]> By: Brian Garrett http://blogs.cisco.com/datacenter/the_power_of_a_san_switch/#comment-10370 Brian Garrett Thu, 22 May 2008 19:35:30 +0000 http://blogs.cisco.com/the_fallacy_of_wire_rate_switching_in_the_data_center#comment-10370 Brian Garrett, the author of the ESG report mentioned here, with a couple comments:1. ESG stand for Enterprise Strategy Group, not Enterprise Storage Group. We changed the name a couple years ago when we started covering other industries including security. 2. Although ESG is well known and respected for hands-on validation testing, the power and cooling data presented in the report is based on technical review and analysis, not hrough actual testing”" as stated in the blog.”

   0 likes

]]> By: Douglas Gourlay http://blogs.cisco.com/datacenter/the_power_of_a_san_switch/#comment-10369 Douglas Gourlay Thu, 22 May 2008 19:35:30 +0000 http://blogs.cisco.com/the_fallacy_of_wire_rate_switching_in_the_data_center#comment-10369 Power and cooling really are the major top of mind issues in the data center today. It seems as if the data center operators have an almost illegal-substance like addiction to compute capacity. I think this breeds an interesting problem-The more work we (as an industry) do to improve the power and cooling efficiencies in the data center the more the operators and administrators will take these efficiency gains and use them to improve the compute density in a given environment.As one friend from a large and storied investment bank told me once, at 70% of physical capacity of a data center we are at 100% of power capacity and our local power company cannot give us more”".I was in a roundtable with Andy Karsner, the Assistant Secretary of Energy, several weeks ago and the entire discussion was centered around data center efficiency issues and how the industry could partner with the government to jointly help. Some of the key points raised at this event were:1) A move to a more efficient voltage than 110v. This could yield significant efficiency improvements without requiring new capital expense for the IT assets. (credit here goes to Andy Bechtolsheim from Sun).2) A common set of efficiency guidelines and measurements for power supplies at different load thresholds. (this is primarily a concern for the server vendors since low-cost white box vendors often skimp here and some power supplies have been seen that rate as low as 50-60% efficiency at nominal loads. This would be ‘Energy Star-esque’.3) The ‘field of dreams’ reality that anything we do to improve efficiencies will not result in power reductions but will result in increased capacity.4) One gent commented that we should raise the operating temperatures of silicon chipsets. That by raising the optimal temperatures from 68F to 70F or so would allow them to run the HVAC less. While at a cursory glance this seems to make sense CPUs take more power at higher temperatures to operate and this would not be a win.5) Server ‘sleeping’ and Virtual Machine integration provides some interesting opportunities- could we move virtual machines to a server in the evening and have a ratio of say 50 or 100 Guest OS’s to 1 physical server and then in the morning move the guest OS back to a dedicated server at a 1:1 ratio.This would allow the servers to use IPMI to turn off at night and Wake-on-LAN to come back online in the morning. 6) The last thing that came up that I found very interesting was that when it got over 110 in August in Northern California we were 385 Megawatts from rolling blackouts in Norcal. Doesn’t really click home for me until I remember that we have some customers building 100MWatt data centers now. Thus we were 4 data centers from rolling blackouts in Norcal – and California as a stand-alone is the worlds’s 6th largest economy. Companies are moving facilities (and thus tax base) out of California to the Colombia river valley, to Phoenix, etc because they are not assured of power grid facilities and self-generation is still perceived as too complex/expensive.While we have created in the networking world data-plane virtualization technologies to reduce the overlap in SANs, more efficiently utilize ports, etc we have not started control-plane virtualization yet on our products. I think the move to integrating control plane and data plane virtualization together between networking products and compute and storage devices will provide interesting new opportunities to lower the power and more efficiently use what we have. Although as to whether we take the energy efficiency gains and return them to the bottom line or increase the compute and storage capacity of a fixed facility I think that is a larger societal issue… What would you do?”

   0 likes

]]>