Archive for July, 2008

Hot Computing Going Environmental Green in the Extreme

Virtualization Critical Evaluation – Chapter 04

This article is a milestone for a couple of reasons. First, a Proper Virtual World as a blog has been around for year or so. Although popular to a reasonable and rational extent, there have been several high notes, where specific articles have been extremely popular, threatened, hated, if not described by every negative adjective know to human kind. At times I believe VMware hates this blog, EMC shakes its collective head in demur surprise, and Microsoft decides that stepping on this ant is not worth the effort to do so, well, fine. In defense, I do know that there are some that love this blog, because I have said things that make people think, question, and re-evaluate their perspectives, and that, is goal of this author. Second, this article is unique because it is not talking about virtualization, but about the infrastructure that supports virtualization. Energy is a significant, critical resource, and the computing industry has done, as a whole, from my perspective, a horrible job reducing energy consumption. So this article is going to discuss why and how we, the information technology industry, as a whole, should change its use of energy.

We need to agree or establish one key point for common understanding, heat should not be feared, heat generation is not bad in of its-self, but carbon generated or released into the atmosphere is. Why is this important distinction made? I will explain further in this discussion. But for now, let us explore what has been done? Reduction of total infrastructure, and the corresponding reduction of offset cooling, is reduction of energy use. For example, if we are talking about datacenters, anything that reduces infrastructure cost is a step in the right direction:

  • Virtualization reduces total infrastructure, reducing heat, cabling, total servers purchased, etc., thus reducing heat generation, and energy needed for cooling.
  • Processors, dual-core, quad-core, etc., have reduced total servers purchased, processors can now step down or up based on processing demand or load, reducing energy needed for function.
  • Centralization of computing resources. This is a work in progress for many entities but it is placed here, small sites, regional datacenters incur infrastructure costs that with modern networking speeds, 10GB or better for Ethernet, and 8GB or better for fabric, allowing for re-centralization, should not be incurred.

The above successes beg the question, why is offset cooling still used to such an extensive degree? Reduction of heat by design is not the same as management of heat, thus offset cooling is handling heat that has already been generated, because this heat is feared, since it will damage or reduce the lifespan of equipment. The goal should be survival of heat by design, making the need for offset cooling as irrelevant as possible. Now what has not been done? What could have been done better? What should be done moving forward?

  • Everything should be fiber based. Retire copper transport completely, everything based on fiber, but it switches, routers, backbones, etc. Why not. This makes sense. Photons which transverse glass/plastic, generate less heat than electrons pushing through copper. Electrons movement is a very inefficient chemical and physical process that generates heat. Photon migration with in fiber reduces the chemical and physical interaction to the lowest level possible.
  • Development of more realistic direct-power backbone concepts? My personal experience with direct-power models for computing has not been extensive, but discussing this topic with friends and experts this technology is not living up to its potential. The best I have seen is a 5 or 10 percent total power reduction, and to get near 15% an entire datacenter has to be built with this technology in mind. That is a hard position to take with existing datacenters that can not be converted, but must be abandoned, since entire buildings must be redesigned for this concept to be effective. There has to be something better.
  • More extensive use of the utility computing models, that are application instance based. Operating system isolation demands millions of lines of code run, plus intensive use of redundant resources, for RAM, Disk, etc.. Whereas application instance based virtualization model reduces total lines of code execution foot print, which reduces the need for faster, bigger energy hungry processors as well as the need for the redundant infrastructure thus reduceing heat generation, and need for offset cooling.
  • Development of computing at above room temperature? This is the trick, just like cold fusion would save the world. A hot computing model, this is a great idea, but it has gone almost no where? Processors that use heat, not avoid heat? Processors that do not generate significant heat? Photon based processors? This is not science fiction, or at least is would not be, if we committed to its development. This should have a bunch of physics graduate students going crazy at 3am in the morning across every University in the world working on this idea, no?

True, we have some technology that works without offset cooling. For example, all our mobile communication and computing options achieve this. However they are also some of the most inefficient devices in reference to energy consumption, and battery technology lags. The latest battery film isolation materials may break through and impact this trend. But practical application of these new battery films is still pending.

Remember, I said heat generation should not be feared, it is atmosphere carbon that should be? The Earth survives, or more to the point, life survives because the Earth is warm. Space is cold, very cold. In fact, the Earth gives off a significant amount of heat, no problem. Carbon, maybe the most significant offender, in the atmosphere blocks this heat loss, and so global warming, such as it is, regardless of the source of carbon, occurs. So that is why I said, generation of some heat is not or should not be a problem. It is how we generate the power and how we use power that generates Carbon, which is the problem. We can generate all the heat we want, and as long as we do not trap the heat in the process of generating the heat, thus we will avoid global warming. Nuclear power is a perfect example of this, and why I believe in Nuclear power generation. It generates heat, and leverages heat, but it does not generate the Carbon, that traps the heat generated. So, we can generate heat, we design computing to survive heat, and not design environments that must control heat, such as offset cooling. Taking this idea further, imagine a battery concept that is more efficient for mobile computing, a datacenter model that embraces heat, by design? Now that would be extreme green.

We sent mankind to the Moon, we sent semi-autonomous robots to Mars, we launched objects in near deep space for decades, and some objects beyond our Solar system. But we can not reduce the total energy foot print in the computing industry, and significantly eliminate the need for offset cooling? In fact, energy consumption has seen a dramatic increase because of integrated computing, in just about every way you can think of, be it large appliance to hand-held device. What is wrong with this picture? Even if energy costs are not up by 50 or 100% at times, why have we not reduced computing infrastructure energy use by 50 or even 75%? Yes, elimination of the need for offset cooling would just about do it, no? Is not your summer air conditioning bill, about 50 or 100% more than your average winter bill? Well, mine always seems to be!

Add comment July 30th, 2008

Does My Dachshund Know More About Virtualization Than You?

Virtualization Critical Evaluation – Chapter 02

My dachshund, well Dachshunds, I have two actually, one is quite old, 17 and wise, in dachshund terms, where the other is younger, but not a puppy, 11 years old, and even smarter. At times I think either of them has more common sense than most strategic theorists in the information technology industry. No, am not talking about Gartner, although I can understood why you would think that at first. Gartner as an organization has more individuals that can state the obvious today, but guess at the future, than I could believe anyone could have. But as I said, I am not talking about Gartner. No, I am not talking about VMware. Buzz, buzz, wrong, buzz, wrong, but thank you for playing. Please play again. Anyone guess Microsoft? I am, talking about Microsoft. Yes, Microsoft.

If there was a Raspberry award (Razzie) or something similar for the information technology industry, oh, let us say, the Frozen-Or Just-Again Reboot (FOJAR) Award, then Microsoft would not only receive just about every FOJAR, in just about every category, but would of course anyone not be surprised when Apple Computer was the most significant sponsor. I can just see the trailers for the show now… Watch the FOJARs, on iTunes! Of course the most memorable FOJAR won by Microsoft this year was for the most significant missing feature in a modern hypervisor, in Hyper-V, the not-so-transparent-almost-not-really-real-time virtual instance migration! Talk about the power is on, but no one is computing!

Why did Microsoft after more than two(2) years, no three(3) years, release Hyper-V without the single most significant feature that everyone doing virtualization is chasing? A feature that is 100 percent identical to VMware VMotion? Even my Dachshunds know this was not a good idea. In fact, it is little better than a joke among virtualization architects that I know. But, I think I understand why it happened, Microsoft is afraid of looking like they are standing still compared to VMware (http://news.cnet.com/8301-10784_3-9980571-7.html, not the article, in the comments on the same page… from Penguinisto is classic). Another reason, which just makes the perception of Microsoft is in fact doing little more than standing still in the hypervisor market, is that Microsoft has completely lost its ability to innovate?

Yes, I know others have said this before, but it has never been more obvious than now, true? I see little improvement in Hyper-V to Microsoft Virtual Server 2005 R2, or even Connectix Server which Microsoft purchased, what, some five(5) years ago. The real surprise is that Gartner has not said this at least twice in 2008, nor noted it as a strategic fact, cough, prediction for 2009? Talk about missing the obvious?

Of course, not commenting on Microsoft System Center Virtual Machine Manager (SCVMM), would be a mistake on my part, a fact of which both of my Dachshunds have just reminded me, would be as unforgivable was running out of dog chews over the weekend. Last time we, I mean I, failed to immediately go the store and restock up on dog chews, you would not believe the dirty looks I got from my Dachshunds. And as any one that has a Dachshund knows, Dachshunds are masters of the dirty look, that-do-it-now-or-else no nonsense stare. But I digress. Obviously Microsoft is ignoring the stares from those of us that love virtualization? Maybe not completely? As I have said before, Microsoft has a true threat to VMware with SCVMM, aimed at the VMware strategic flagship, VirtualCenter. However, I have to disagree with my esteemed Dachshunds, SCVMM without a VMotion comparable feature, read comparable as transparent to end-user migration? Never mind Storage VMotion? And if Microsoft SCVMM does not scale better than VirtualCenter? Well, Microsoft still gets two FOJARs, the first for Hyper-V, the second for SCVMM if it does not nail VirtualCenter to the wall. No, Microsoft gets three(3) FOJARs. Why? Well, Microsoft gets the hat-trick FOJAR because they have taken more than 60 months to go almost absolutely nowhere in the fastest era of virtualization adoption the information technology industry has ever seen.

Gartner, how in the world did your crystal ball miss this one? Maybe Microsoft will create time travel? So Microsoft can innovate in the future, but release to the market today? No, we already established Microsoft does not innovate. Hey, maybe they can buy time travel from someone in the future… Yeah, that is it!

Add comment July 16th, 2008




Calendar

July 2008
M T W T F S S
« Jun   Aug »
 123456
78910111213
14151617181920
21222324252627
28293031  

Posts by Month

Posts by Category