How do I personally define the cloud? At the risk of over-simplifying: Online banking. Think about it: I ultimately run my business based on information. Information about my products, my customers, my employees, my finances, etc, and I need that information to be presented to me in such a way that it enables me to make decisions and to take (or not take) action. Online banking works just like that. It presents information about my account balances, gains/losses, etc. and based on that information I decide if I need to save more, if I can afford to spend any, if I need to move some around, etc.
Ultimately I only care that I can get to the data by calling my banker, or preferably, via a computer, netbook, terminal, or smartphone with a browser. I don't care where the data sits as long as I can get to it and it is secure. I don't care about the application that serves it up to me and I certainly don't care about the operating system or hardware on which it runs. I'd bet that most CEO/CFO types would tell you that this is exactly how they feel: Give them access to the data they need, wherever they are, whenever they need it, and make sure it is secure at rest and in transit.
Tuesday, March 16, 2010
Cloud services drivers?
From an executive perspective, I think we have been waiting for "the cloud" for some time now. Ever since the advent of the PC and it's insertion into the corporate environment, we have been trying to figure out how to manage users and data.
In the mid-90s, I was involved in rolling out IBM java-based thin-clients as terminal replacements hanging off of their mid-range platforms. This was a small step in the right direction (IP-enabled, Ethernet-attached, GUI-based) but still just a terminal with no real ability to match the performance characteristics or application availability of the PC, so it died on the vine.
Around the same time, we saw a hint of things to come with the Application Service Provider. Sure, people were tired of paying over and over again (capex) for updated versions of software that gave them no real advantage. Most people just want to create simple documents. Did each new version of Word make us that much more creative and powerful? The problem at that time was a lack of affordable bandwidth to deliver those apps as a service.
Then we saw the telecom bubble expand and miles and miles of fiber were laid to tap into the greatness that the Internet was quickly becoming. Then that bubble burst, 9/11 happened, and things shrank back from the bold new horizons that we looked to at the time. Simultaneously throughout the late '90s and into the early 2000s, we were hearing that we should focus on our core business and that IT was "context" and just a burden and while we understood where that was coming from, there was really no alternative.
And, finally, during that time, the last piece of the puzzle was slowly emerging: Virtualization technology.
So now, here we are. We have the mindset, the bandwidth, and the virtualization technology that allows us to decouple our data and applications from hardware which removes geographical limitations. We have all of the pieces to build out the Cloud 1.0. There are still a lot of questions around access, security, and offerings, but we are at a point where widget manufacturers can start to focus more on manufacturing widgets and begin to outsource those (IT) things that have ceased to provide a competitive advantage but that are still a necessity to function.
I think Nicholas Carr in his book, The Big Switch, does a great job of explaining why most IT functions will move to a utility type of model. He likens it to electricity. In the early days of electric power, companies had to build their own power generation capabilities on premise. Then they had to staff those systems to ensure that they were operational. One day, someone came along and offered them an outlet into which they could plug in their machinery and only pay for what they used, as they used it. Now the textile mill could get out of the electricity business and leave the generation of that power to someone that could do it on a much greater scale and provide it at a lower price than they could generate it for themselves. He sees the core functions of IT (computing and storage) as commoditized utilities that can easily be "generated" and delivered to any company that needs them and ultimately for a lower cost.
If companies can accomplish their business objectives more cost-effectively by pushing a bulk of their IT operations out to a cloud of some sort, then I am betting they will. After all, as Nicholas Carr says: [In] the end, the savings offered by utilities become too compelling to resist, even for the largest enterprises.
In the mid-90s, I was involved in rolling out IBM java-based thin-clients as terminal replacements hanging off of their mid-range platforms. This was a small step in the right direction (IP-enabled, Ethernet-attached, GUI-based) but still just a terminal with no real ability to match the performance characteristics or application availability of the PC, so it died on the vine.
Around the same time, we saw a hint of things to come with the Application Service Provider. Sure, people were tired of paying over and over again (capex) for updated versions of software that gave them no real advantage. Most people just want to create simple documents. Did each new version of Word make us that much more creative and powerful? The problem at that time was a lack of affordable bandwidth to deliver those apps as a service.
Then we saw the telecom bubble expand and miles and miles of fiber were laid to tap into the greatness that the Internet was quickly becoming. Then that bubble burst, 9/11 happened, and things shrank back from the bold new horizons that we looked to at the time. Simultaneously throughout the late '90s and into the early 2000s, we were hearing that we should focus on our core business and that IT was "context" and just a burden and while we understood where that was coming from, there was really no alternative.
And, finally, during that time, the last piece of the puzzle was slowly emerging: Virtualization technology.
So now, here we are. We have the mindset, the bandwidth, and the virtualization technology that allows us to decouple our data and applications from hardware which removes geographical limitations. We have all of the pieces to build out the Cloud 1.0. There are still a lot of questions around access, security, and offerings, but we are at a point where widget manufacturers can start to focus more on manufacturing widgets and begin to outsource those (IT) things that have ceased to provide a competitive advantage but that are still a necessity to function.
I think Nicholas Carr in his book, The Big Switch, does a great job of explaining why most IT functions will move to a utility type of model. He likens it to electricity. In the early days of electric power, companies had to build their own power generation capabilities on premise. Then they had to staff those systems to ensure that they were operational. One day, someone came along and offered them an outlet into which they could plug in their machinery and only pay for what they used, as they used it. Now the textile mill could get out of the electricity business and leave the generation of that power to someone that could do it on a much greater scale and provide it at a lower price than they could generate it for themselves. He sees the core functions of IT (computing and storage) as commoditized utilities that can easily be "generated" and delivered to any company that needs them and ultimately for a lower cost.
If companies can accomplish their business objectives more cost-effectively by pushing a bulk of their IT operations out to a cloud of some sort, then I am betting they will. After all, as Nicholas Carr says: [In] the end, the savings offered by utilities become too compelling to resist, even for the largest enterprises.
Monday, March 15, 2010
VDI and Storage (briefly)
The storage subsystem may be the most important piece of the entire virtual desktop infrastructure. It is responsible for most of the performance of the virtual desktop environment and if it can’t keep up, the entire environment will be impacted. We learned this the hard way a few months back.
In this environment, the I/O per second (IOPS) capacity is critical. Mechanical drives (your standard spinning disk) just can’t keep up with the demands a high-density VDI puts on them. Products you should consider are built around solid-state memory and provide thousands of IOPS per GB vs. just a couple hundred per GB with mechanical drives. Yes they are more expensive per GB but they are a fraction of the cost when you break it down on a $/IOPS basis.
In this environment, the I/O per second (IOPS) capacity is critical. Mechanical drives (your standard spinning disk) just can’t keep up with the demands a high-density VDI puts on them. Products you should consider are built around solid-state memory and provide thousands of IOPS per GB vs. just a couple hundred per GB with mechanical drives. Yes they are more expensive per GB but they are a fraction of the cost when you break it down on a $/IOPS basis.
VDI and the trouble with anti-virus
I don't know if this is news to most folks but hopefully it will provide some understanding of the problem of running traditional anti-virus software in a virtual desktop environment. This comes from experience...
Anti-virus is a definite concern in Virtual Desktop Infrastructure (VDI) deployment and no one that I have found out here in the blogosphere or on the vendor sites recommends it (except for the AV vendors). The problem is the way AV functions in the virtual environment. The people that don’t believe it causes hang-ups don’t seem to understand what is going on in this environment.
In this environment, the desktop is, for all intensive purposes, generated from a single, pre-built image each time someone boots up. The problem with desktop AV is that whenever it comes online, it goes out to see if it is up to date. When a virtual desktop spawns from the image that includes AV, it is always whatever version you built into the image (let’s call it v3.0). So each time one or one-hundred virtual desktops are spawned, that v3.0 AV tries to “phone home” to see if it is up-to-date. Of course, it is not up-to-date because you built your image five months ago, so then it (and any other image that just spawned) starts pulling down all of the updates to get it up to the current rev of v3.8. By the way, in our non-persistent environment, as soon as I log that desktop off it is "destroyed" and all of those updates disappear so that when I bring up my next instance, I go right back to v3.0.
Now in a physical environment automatic updates are no big deal other than the fact that they may impact the Internet connection when they all try to suck those updates through the straw. In a virtual environment, all of those images reside on the storage system and so every one of those disks is spinning trying to make these virtual desktops appear to operate like a physical desktop and that’s when you get into trouble.
To date and from what I have seen, there is no AV solution for the virtual desktop that is analogous to what we have used in the physical desktop world. That is one reason why people deploy virtual desktops in a non-persistent manner: If I get a virus, no big deal. I’m going to destroy that entire desktop image and everything with it when I log out of it. What if it infects the files? There are AV solutions for the storage that can go and scrub everything on the storage system where we keep the user files and everything else.
So, all of that being said, the AV folks are furiously working on a solution that works here. Just keep this important element in mind as you move down the VDI path.
Into the fray
Well, on my final day of employment at one place and in honor of new (ad)ventures, here is the introductory entry into my new blog site. Since anyone can show up in the blogosphere with their own creation why wouldn't I want to do the same. I hope there will be valuable introspection in the things that I publish around cloud computing from the perspective of the utility model.
I will post non-technical, managerial/executive-level perspectives of things going on in the world of cloud/utility computing and various posts based on my experiences as a senior manager of professional services engineers involved in developing best practices/methodologies and delivery of those technologies, having been involved in "troublesome" projects, and in my new ventures.
For more info on me, feel free to check out my LinkedIn profile.
Subscribe to:
Posts (Atom)
Posts
