Gordon's Blog

coming to a conference near you...

Yes, your friendly neighborhood global technology evangelist is hitting the road and going on tour! I'll be speaking at a number of events hosted by our friends at Datacenter Dynamics as well as several Green IT conferences in the US and Europe. This is very exciting as there is nothing better than meeting with  architects, technologists, developers, IT administrators, and business leaders and discussing with them their unique business needs and challenges.

I was in Tokyo last week at the Datacenter Dynamics "Command and Control Summit" where I had the opportunity to speak to the attending delegates about global-scale utility computing, and how to get there by leveraging a dynamic application services management platform. I found it most interesting, given the very conservative approach that most Japanese businesses have to adopting new technologies, that the delegates were very interested in the Green IT benefits and consistent QoS that can be realized by evolving from a static to a dynamic resource allocation model.

The sushi was also quite good :)

Next up:

If you think you might be in London next week, you won't want to miss my presentation "Going Green: The Shift to Dynamic Application Service Management" - no spoilers here, but I am confident that you'll be green with envy (chuckle - I couldn't resist that one...) if you miss it. And if you do make it, please make sure that you introduce yourself and let me know what your company is doing in order to Go Green.

Keep your bot (or your eyes if you are a tad old fashioned) on this spot and I'll be updating with additional conferences and dates. And if we should meet at a conference I'll be sure to give you a shout out right here :)

Green utility cloud computing

<chortle>

I couldn't resist - everyone seems to be doing it, and I figured that if I were going to jump in, I might as well make a big ol' splash :)

Don't worry, though, I'm not going to muddy the waters with yet another definition of what "cloud" computing is, or extol the virtues of how our software is "green", or agile, or non of the above. Instead I think that I'd like to come out and state for the record that NONE of these high-hit-rate buzz words, or their corresponding concepts, will flourish in your grand-daddy's data center. None. Nil. Zip. Nada.

Oh sure, everyone in the software infrastructure world is going to have an environment saving, agility increasing, cost reducing green utility cloud for SOA and SaaS computing, and some of of these are going to be quite elegant. And some will be very innovative. And useful. Likely they will all be expensive because, well, green utility cloud computing just isn't cheap, not for all the money you get to save. But as long as IT and the data center continue to manage infrastructure in the same old way they have for the past gazillion years, none of these green utility cloud computing solutions is going to bear the kind of fruit that the buzz-doctors are chattering about.

Why?

Resource allocation. That's why.

Until organizations take an active position in correcting their resource allocation model, they just won't be able to have that shiny new green utility cloud.

So what's the problem with your allocation model?

It's static (and manual), that's the problem. Only it doesn't look like a problem because everyone is doing it. It's the way it's always been done. Static allocation is the reason that infrastructure utilization is so low, and ironically, why virtualization has become so popular. I say ironically because all you're doing with virtualization is creating more resources - increasing the supply - only to turn around and statically allocate them - just like you always do. Yes, I know, you're driving utilization and reducing the amount of physical servers, which is all great news until you've gone and statically allocated all of those virtual resources. Then what? You've got idle virtual resources, a lot more management and complexity, and you are still looking for more resources because you have more demand, more applications, and increased qualities of service.

The solution is to make your allocation dynamic (and automated) - that's the easy part. The tough part is the fundamental - paradigm - shift that must take place in order to deliver dynamic, automated, demand-based allocation of resources to the applications that need them, when they need them, only for as long as they are in need, and all based on priority. It's tough because it means changing the way you do business today. It means that IT needs to become a Quality of Service provider, not a caretaker of someone else's servers. It means that the LBUs no longer get to say how many servers they need, instead they pay for a consistent quality of service, and leave the allocation of resources to accommodate their requested QoS to the folks in the data center.

It means change is necessary if you want your very own green utility cloud.

Change is good.

Tags: green, utility, cloud, resource allocation, grid

Treat the illness, not the symptoms

Organizations are like most individuals when it comes to suspicions that something is wrong with them: they put off the trip to the doctor rather than be given some potentially bad news. Instead, they continue to treat the symptoms and refuse to admit that they are likely not treating the illness.

In this case, I am speaking of an illness that is so widespread that it affects virtually (err, no pun intended) every organization - yet very few have realized it and begun to take action to cure the illness. The illness is "The Way We've Always Done It" Paradigmitis.

"The Way We've Always Done It" Paradigmitis is likely crippling your business right now. You see it and feel it every time you walk into the data center, read a utilization summary, or see your energy costs: resources are grossly under-utilized, applications are over-provisioned for peak demand, power consumption and costs are increasing, and you are running out of room and time. These are the symptoms of the illness; however they are not the illness itself.

The illness is an inability of the IT organization to function efficiently and optimally by providing a real-time infrastructure to business units. It is the direct result of separate operating and business units within an organization telling IT what they want instead of contracting for what they need. It is insidious because it is universal, has been around since the dawn of the data center, and, like a chronic illness, we have simply learned to live with it.

So, in our denial, we treat the symptoms. Under-utilized physical resources (servers) are doubled, tripled, quadrupled as virtual resources - more supply is created from existing resources. Applications that were suffering from performance or availability issues now have a larger pool of (virtual) resources to draw on, and these resources are statically allocated to applications based again on what the business unit says it wants. IT has taken a pill and treated the symptoms, but the illness is still there- growing more virulent and insidiously positioning itself for a massive infection of the entire enterprise.

When the effects of treating the symptoms have worn off, organizations will see that they are still quite ill. Only then the illness will present itself in the form of 100s or 1000s of idle physical and virtual servers, business units that want even more resources, and IT is right back to buying infrastructure - attempting to treat the symptoms. At that point, the symptoms will be too acute to treat with the same bandage being used to cover the bigger problem.

You can take action today and confront the underlying issue:

1. Realize that the data center doesn't have a utilization problem, it has an allocation problem. Resources are allocated statically at deployment time, and not dynamically at run-time. The resulting infrastructure is “siloed” and inflexible, which leads directly to under-utilization.
2. Realize that the allocation problem is caused by an IT resource delivery model that is predicated on delivering servers to the business based on peak demand scenarios - giving them what they want - instead of delivering a consistent quality of service to the business applications - giving them what they need, when they need it.
3. Commit to changing the status quo - actually do something to address the issues outlined above by shifting the focus from statically managing resource supply to dynamically addressing resource demand.

Call it a voluntary paradigm-ectomy.

The illness is real, and so is the cure. The question is, are you going pop a few pills, or are you going to call a doctor?

(Yes, we make house calls :)

Paying for the baggage

If you don't travel frequently, you probably missed the news that United Airlines is now charging $25 for checking a second bag, and $100 for a third. The reason given for the fee is the cost of fuel, which is the single greatest expense that the airline faces. I travel a fair amount, and I can't say that I am surprised by the move, nor am I put off by it. The business, like people, must adapt to changing environments in order to survive. This is evolution 101.

Interestingly, this is true of today's enterprises as well. The environment (quite literally) is changing and organizations must adapt in order to survive. Evolution or extinction - two sides of the same coin, really.

The signs of adaptation are all around us: virtualization, data center automation, service-oriented infrastructures and architectures, and green IT - all responses to the economic and environmental costs of doing business. However, adapting to one's environment, and using that adaptation effectively are two different things entirely - just ask the Neanderthals.

Organizations are either going to have to get smart about how resources are allocated and utilized, or they are going to have to start paying for their baggage too.

a rose by any other name, is still a rose...

The word is out - FabricServer 2.5 is released - and the name of the rose is Real-Time Infrastructure. While this may be a relatively new term, the concepts and IT models that Real-Time Infrastructure encompasses have been around for quite some time:

• utility computing
• on-demand computing
• service oriented architecture
• software as a service
• shared services

All of the organizations and people that I speak to are, in some way, moving in the direction of Real-Time Infrastructure, and they are doing it for a number of reasons:improved service-levels and application performance, increased utilization of resources, and reduced opex cost and complexity. They may not be calling it Real-Time Infrastructure, but that's really what's going on when resources - servers, network, storage - are allocated dynamically to the execution of applications and services based on policies which define the relative importance of the consumer and/or that which is being consumed.

It is important to point out that Real-Time Infrastructure shifts the relevance of virtualization from simply increasing the supply of resources (a la the OS virtualization technologies) to automating the response to the demand for these resources in a manner that ensures that applications and services have the resources they need, when they need them.

Are you building a Real-Time Infrastructure? If so, why not tell us about it! And if not, what's stopping you?

2008: The Year of... What?

Green Computing? Grid Computing? Virtualization? SOA? These aren't the only topics being discussed as relevant in 2008, however they are the topics being discussed the most. Which of these, in the next 11 months and 29 days, is going to impact you and your business?

For me, the answer is "E - All of the Above" :)

Not because they are topical and worthy of thought or discussion or implementation (they are!), but because they are all responses to a common problem manifest in today's data centers: utilization. For as long as there have been data centers supporting the business, the approach typically taken to ensure the *-ilities - availability, scalability, reliability - of business critical applications has been to over-procure and over-deploy hardware in static, siloed, under-utilized configurations.

Clearly, continuing to to do things "the way we always have" isn't an option, not any more. So, here's to 2008: The Year of... Doing Things Differently!