Archive for November, 2007

Charge Back Models? A Lack of Financial Brain Power?

Know What Virtualization Is, But What Is Next? – Chapter 06

An interesting issue, at least for me, keeps coming back to the surface of issues that we, as virtualizationists have to deal with. Oh, virtualizationists? Well what the heck do we call ourselves? We save tons of money by implementing virtualization, we reduce infrastructure and in effect help make the computing industry more green, well, greener than ever, and I am not referring to the color of the ink on paper money of course. Unfortunately, accountants, the most evil profession on the planet in my humble opinion, just hate the very concept of virtualization. Why do I say this? It has been my experience that this is true. I have sat in various meetings where once I recommend pre-provision of infrastructure to allow virtualization to be responsive and adaptive to business needs, the financial gurus groan, grunt, hum and haw, and in at least one instance, a swear I heard someone crying at the other end of the long conference table, cough, call.

A project driven model, which accounts love, is a nice, specific, limited costing analysis, which is in effect a point-of-sale transaction, for example, the project is determined, management approves it, it is funded, and then implemented, and other than the depreciative costing over time, is a very simple capital expense model. However, pre-provisioning of virtualization infrastructure is not such a simplistic animal. Now, hardware is split by some crazy criteria, which may not even be obvious at first. What will the customer base really use? Virtual CPU capacity, Disk IO, Network IO, or even Memory Loading? What if your costing model is based on just one or two capacities, but all your clients end up using more of one of the others you did not base your cost model on? How in the heck do you remain responsive to customer needs over time?

Even worse, after you figure out some costing factor for Virtual CPU, Network, Disk, and Memory, did you really pre-provision enough or each? What about the time factor? Whatever you did pre-provision the accountants want charged back immediately, but when have you even had 20 virtual instances, or more online the same time you racked the virtual host server? Network connectivity must be preconfigured, who gets that fixed cost per month until it is fully loaded? Shared storage the same issue, but you have to have network and disk resources available before the virtual host is online, right? This is getting complex and difficult to manage from a project or logistical perspective, now they are yelling about cost allocation and charge-back? Some one or somebody save us, please.

Now enters the peanut-butter principle. Never heard of it? Me either, but one enlightened financial guru coined the term in a meeting some time ago, and it has stuck with me ever since. Now I am suggesting it to everyone else as a concept to follow. Here is how it goes:

  • Decide what you need and when you need it in reference to infrastructure, pre-provision at least what you can for the first year if you can, this is a sunk cost.
  • Decide how or what you will charge against, again by virtual CPU, disk, memory, and/or network resources, include all costs for the period defined, site costs, infrastructure costs.
  • Define a charge model, that includes the typical virtual instance, the average scale per instance, divide this value by the total pre-provision cost for the first year, this is an estimate, and will change after the first year, but it does give you an argument to judge virtualization savings.
  • This model is a leap of faith for two reasons, one, you expect to have virtualization for at least several years, and two, you client base understands that the initial costing quote is an estimate only.

There are some issues with this model, first year; it overcharges early adopters of virtualization, and under charges late adopters of virtualization, until a financial correct is done. However, if you bill by resource, and not by virtual instance, you can narrow the over and under costing variance per month. At the end of the year, the client will get credit or demand for funding depending on the baseline accuracy and variance incurred. The peanut-butter factor is the cost incurred per resource, per instance, per client, against the baseline costing. How does the financial team spread the variance and incremental costing that always comes up?  Good question!

Spread, peanut-butter? Get it? For example, 1 VCPU, 1GB RAM, 30GB DASD, with an average 100MB bandwidth need is $500 per month. But the client needed 2GB RAM after 3 months of service, but we had to pre-provision 32GB RAM in the virtual host. And, said client virtualized in January, and our other client virtualized in June. In December, what do we charge the January client, versus the June client? Oh, and did I mention that we had to increase total RAM in the virtual host to 48GB level in October to met demand needs? You really do need away to even out the costing to all clients effectively and accurately.  I have yet to find a great tool or method to do this, but I am sure someone out there has it on their respective roadmap.

Got a headache yet? Now that I think about it, the one accountant that I thought I heard crying? I did not understand their panic attack then, but now, I think I do. Whatever is done to address this issue, it is complex, beyond simple accounting spreadsheets, and will never work great from day one, it will have mature, just as virtualization is maturing. With each push of the knife across the slide of bread, the peanut-butter smoothes out, nice and even. Wait…Ok…Who is the wise ass…? I said smooth peanut butter, but why do I only have a jar of super chunky? Now I want to cry.

2 comments November 28th, 2007

The Halo Trilogy and VMware Have Something In Common?

More than Combat Evolved To Be Sure!

Yes, this title is a bit different from my usual rant, rave, and cursing of the virtualization ups and downs. But since life is a series of metaphors that create perception? Why not a videogame represent reality? I want to explain something that occurred to me, while, yes, playing Halo 3, on my Xbox 360, and no this is not a promotion for Microsoft game console line, or videogames for same.

Suspend reality with me for a few moments and consider the following… In dimension of sight, of sound, of reality not so far different from ours… cybernetics are a real, creation of super humans for combat is not common, but done, Combat Evolved, if you will. Specific to our discussion, a lone survivor of an elite unit, call them Spartans, is fighting to save not only himself, but his race. He is against impossible odds, under scaled resources; even some of his own peers give him no chance of success, in fact they even hate him for his past efforts. This solider, sometimes called the Chief, has only 2 goals, one, save his race, and two, save his friend, an A.I. entity. Cortana, if she, yes she, was human might be his love, rather than just his guide on this adventure. Comments like… Finish the FightWe will… are the few words and not much more that the Chief says. His heart is given to his goals, he acts, he responds, he improvises, he overcomes, and most important, from a videogame perspective, he wins. Like, all the Greek warriors of the past, burned, carved, polished, programmed, down to a lean and efficient core of ability. He was a rumor, then a hero, now a legend.

For those that do not realize or know, what I have illustrated is in basic elemental terms the premises of the game story line of Halo, and its sequels Halo 2 and Halo 3. How John, designated unit 117, a Spartan class combat cybernetic solider was/is introduced to millions of avid XBOX and now recently, XBOX 360 fans over the last 6 years or so. Just maybe the most popular video game of recent history.

Returning to reality, and the explanation of the title of this article, VMware is at a cross roads of the future, down one path, say the left, VMware disappears into the shadows as Netware has done. To the right, avoiding the darkest shadows of market share, finding a niche scenario such as RedHat has achieved. And of course the obvious path, the one straight ahead, the domain of Microsoft, the endless resources of the empire of code, marketing, capital, and even more to the point, attitude. To jump back to the Halo story line, Microsoft could be seen as the Covenant, an alien society that is dominated by a prophetic vision and desire to change the universe to its own belief of reality. Yes, I am avoiding the obvious Star Trek comments, we are discussing Microsoft and Halo universe after all. Back on point, Microsoft is allowed to do this, this is what a free and open market mechanism is, and I support it.  However, this also means that VMware should expect real competition, and when it comes, it will come in force, with massive impact. As in Halo, the Chief has no where to go but up-hill, this is true for VMware as well.

Microsoft just released System Center Virtual Machine Manager (SCVMM), it is not perfect, in fact I find it hard to use, its GUI seems odd and disjointed, and it is not elegant or as sophisticated as it will some day be. However, I do see its potential, it will provide Microsoft with a tool, no, a weapon to attack VMware at the core of its more mature virtualized product line, VMware VirtualCenter. VMware developed a great hypervisor, it dominates the performance of virtual instances, VMware ESX host, although not perfect, is better than anything else I have seen or used to date. But what really catches your attention is VMware VirtualCenter. VirtualCenter makes managing all your hosts at least as painless as possible, it is not perfect, and suffers from VMware growing pains, but it is the best of breed right now, in spite of its quirky performance and lack of scaling to a true corporate enterprise level.

Microsoft does learn and has learned from VMware, it adapts, and this is the danger for VMware. Again, no Star Trek comments please. Although I have yet to push System Center Virtual Machine Manager to the same scale as I have VMware VirtualCenter, I will not be surprised when it scales to a reasonable level for enterprise customers, or that it may only be an incomplete match against VirtualCenter, but that will be enough to rock VMware, and shake it to its very bones. Not to mention painless integration to MOM. Just like a plasma grenade that is a bit to close to the Chief. Can you feel the game console controller vibrate and rattle?

As at the end of the Halo 3 plot-line, picture a scene, in a dark spaceship, lost and a drift, with just two souls aboard. Look into a chamber of shadows, with the A.I., a now rescued Cortana, looking fondly at John 117, as he locks his pulse rife into a nearby weapons rack, then turns and climbs into a stasis chamber. Cortana says something to the effect…I have sent out a beacon, but it maybe years before they find us. To which, the Master Chief replies with his simple, minimalistic style, more his personality than habit… Wake me… (Insert a real pause for effect)…When you, need, me. I am sure the ironic tone of this is not lost in printed form. When I heard it, it gave me a cool chill down my spine. Really it did. 

Thus, in our reality… VMware wake! Competition is what we need! Challenge Microsoft; create quality, consistency, real best of breed solutions, the feature race is not effective, quality and stability is, after all that is how Microsoft did it, even if in the case of Microsoft Word it took them 11 versions to do it.  Microsoft is in this fight to win, and they have the potential to do so, they have adopted the mindset of Halo 3… Finish the Fight. VMware can not afford to sleep; can not afford to be drifting lost in space. As the Chief saved Earth, someone must save the future of virtualization. There may not be a Halo 4 in our future, only the accountants of Bungie know for sure. But I know this much… VirtualCenter 4 has got to kick some major alien ass, or it is, in every sense of the phrase… Game Over.

7 comments November 14th, 2007


November 2007
« Oct   Dec »

Posts by Month

Posts by Category