One of the most enjoyable blog posts I read over the holiday break came from Reuven Cohen, a longtime cloud evangelist and CTO of cloud management software vendor Enomaly.
In a post titled "Is cloud computing actually environmentally friendly," he points to one of the most perplexing aspects of the claim that cloud computing is green:
The problem is there is no uniform way to measure this supposed efficiency. None of the major cloud companies are providing utilization data, so it is not possible to know just how efficient cloud computing actually is--other than it sounds and feels more green.
He addresses the environmental impact of both building data centers and choosing their power sources. He then discusses consumption:
We now have the ability to run our applications on thousands of servers, but previously this wasn't even possible. To say it another way, we can potentially use several years worth of energy in literary a few hours, where previously this wasn't even an option. So in direct contrast, hypothetically we're using more resources, not less. On the flip side, if we bought those thousand servers and had them running (underutilized), the power usage would be significantly higher. But then again, buying those servers would have been out reach for most, so it's not a fair comparison. There we are--back, at where we started. You may use 80 percent less energy per unit, but have 1,000 percent more capacity, which at the end of the day means you're using more, not less energy.
He's right. How can we claim to be creating a greener world through cloud computing with no data to back it up? In fact, there is some significant evidence that the cloud is encouraging more compute consumption, which would erase much or all of the energy savings that the cloud's increased utilization efficiencies would achieve.
Simon Wardley, software services manager at Canonical, discussed this several months ago in his blog:
The shift of common and well defined IT activities from products to standard components provided as online services should lead to a dramatic explosion in innovation.Standardization always creates this potential.
...Cloud computing is all about providing standard components as services (it's pure volume operations). The problem of course is that we will end up consuming more of these standard components because it's so easy to do so (i.e. in old speak, there is less yak shaving) and it becomes easier to build new and more exciting services on these (standing on the shoulders of giants).
We might end up providing more efficient virtual resources but we will end up consuming vastly more of them.
So, in a sense, the "greenness" of cloud computing is a kind of Schroedinger's box problem today, in which we won't know the actual savings to the environment until someone actually observes--or measures--it.
Krishnan Subramanian, a commentator for CloudAve, is also aware of this paradox, but he points to a story out of Finland that shows some hope that there are additional ways we can help the environment through cloud computing. Apparently, the city of Helsinki is using a data center operation to heat homes--a perfect way to gain value at both ends of the heat exchange process.
I believe one thing to be true: the increased efficiency of the hardware components in most cloud data centers and the increased utilization of these components mean that we are almost certainly doing more work per unit of energy consumed than before. However, I think we'll have to wait awhile before there is evidence of the overall effect of cloud computing on the planet...one way or the other.
In a much anticipated move, Microsoft announced the combination of the Windows Azure group with the Windows Server and Solutions group into a new organization, titled the Server and Cloud Division. The new division, headed by Senior Vice President Amitabh Srivastava, will be a part of the Servers and Tools Business, headed by Bob Muglia.
The new division will "deliver solutions that help our customers realize even greater benefits from Microsoft's investments in on-premises and cloud technologies," according to the Windows Server Division blog. The Windows Azure team blog adds that the combined team will "ensure that customers get the full benefit of Microsoft offerings that span Microsoft's public cloud, on-premises solutions, private clouds, and clouds that our partners host."
The move makes sense, as the company's "software plus services" strategy requires consistency in the management and execution capabilities of both Windows Server and Windows Azure. Microsoft has been working on both Azure and private cloud capabilities for some time now, though its Web site currently pitches its Dynamic Data Center Toolkit as a "foundation" for both private and partner cloud services.
It should be noted that this move means that CTO Ray Ozzie is no longer heading the Azure team, a signal that Azure has graduated from a technical project to a full-fledged Microsoft business.
Also announced was the move of the Windows Azure Business and Marketing team, headed by Doug Hauger, to the Server and Tools Marketing group, under Corporate Vice President Robert Wahbe.
In the opening post of this series, I joined Chris Hoff and others in arguing that cloud computing will change the way we package server software, with an emphasis in lean "just enough" systems software. This means that the big, all-purpose operating system of the past will either change dramatically or disappear altogether, as the need for a "handle all comers" systems infrastructure is redistributed both up and down the execution stack.
The reduced need for specialized software packaged with bloated operating systems in turn means the virtual server is a temporary measure; a stopgap until software "containers" adjust to the needs of the cloud-computing model. In this post, I want to highlight a second reason why server virtualization (and storage and network virtualization) will give way to a new form of resource virtualization.
I'll start by pointing out one of the unexpected (for me at least) effects of cloud computing on data center design. Truth be told, this is actually an effect of mass virtualization, but as cloud computing is an operations model typically applied to virtualization, the observation sticks for the cloud.
Today's data centers have been built piecemeal, very often one application at a time. Without virtualization, each application team would typically identify what servers, storage and networking were needed to support the application architecture, and the operations team would acquire and install that infrastructure.
Specific choices of systems used (e.g. the brand of server, or the available disk sizes) might be dictated by internal IT "standards," but in general the systems that ended up in the data center were far from uniform. When I was at utility computing infrastructure vendor Cassatt, I can't remember a single customer that didn't need their automation to handle a heterogeneous environment.
But virtualization changes that significantly, for two reasons:
The hypervisor and virtual machine present a uniform application programming interface and hardware abstraction layer for every application, yet can adjust to the specific CPU, memory, storage, and network needs of each application.
Typical virtualized data centers are inherently multitenant, meaning that multiple stakeholders share the same physical systems, divided from one another by VMs, hypervisors, and their related management software.
So, the success of applications running in a virtualized environment is not dependent of the specialization of the underlying hardware. That is a critical change to the way IT operates.
In fact, in the virtualized world, the drive is the opposite; to create an infrastructure that drives toward homogeneity. Ideally, rack the boxes, wire them up once, and sit back as automation and virtualization tools give the illusion that each application is getting exactly the hardware and networking that it needs.
Now, if the physical architecture no longer needs to be customized for each application, the question quickly becomes what is the role of the virtual server in delivering the application's needs. Today, because applications are written against operating systems as their deployment frameworks, so to speak, and the operating systems are tuned to distribute hardware resources to applications, virtual machines are required.
But imagine if applications could instead be built against more specialized containers that handled both "glue" functions and resource management for that specialization--e.g., a Web app "bundle" that could deal with both network I/O and storage I/O (among other things) directly on behalf of the applications it hosts. (Google App Engine, anyone?)
A homogeneous physical architecture simplifies the task of delivering these distributed computing environments greatly, as there is a consistency of behavior from both a management and execution perspective. However, as it turns out, a homogeneous virtual container environment has the same effect.
So, if the VM isn't hiding diversity at the hardware layer, or diversity at the software layer (which is hidden by the "middleware") what is its purpose? Well, there is still a need for a virtual container of some sort, to allow for a consistent interface between multiple types of cloud middleware and the hardware. But it doesn't need to look like a full-fledged server at all.
Thus, the VM is a stopgap. Virtual containers will evolve to look less and less like hardware abstractions, and more and more like service delivery abstractions.
In my next post, I want to look at things from the software layers down, and get into more detail about why applications will be created differently for the cloud than they were for "servers." Stay tuned.
Chris Hoff, my friend and colleague at Cisco Systems, has reached enlightenment regarding the role of the operating system and, subsequently, the need for the virtual machine in a cloud-centric world.
His post last week reflects a realization attained by those who consider the big picture of cloud computing long enough.
He summarizes his thoughts nicely at the opening of the post:
Virtual machines (VMs) represent the symptoms of a set of legacy problems packaged up to provide a placebo effect as an answer that in some cases we have, until lately, appeared disinclined and not technologically empowered to solve.
If I had a wish, it would be that VM's end up being the short-term gap-filler they deserve to be and ultimately become a legacy technology so we can solve some of our real architectural issues the way they ought to be solved.
Hoff goes on to note that the real problem isn't the VM, but the modern operating system:
The approach we've taken today is that the VMM/Hypervisor abstracts the hardware from the OS. The applications are still stuck on top of operating systems that don't provide much in the way of any benefit given the emergence of development frameworks/languages such as J2EE, PHP, Ruby, .NET, etc. that were built around the notions of decoupled, distributed and mashable application "fabrics."
My own observation here is that our current spate of operating systems were designed when competitors were pushing to use the OS as a differentiator--a way of distinguishing one company's product experience from another. OSes started out being targeted at software, providing a way for applications to use a generalized API to acquire and consume the resources they needed.
At the time, computers had one CPU and the logical thing to do was to design a single OS that could run multiple applications, preferably at once. This created the need for additional functionality to both manage resources and manage the applications themselves.
Furthermore, the operating system increasingly targeted not the needs of software, but the needs of people; more specifically, the needs of computing buyers. Take a look at OS X, or Windows, or even "enterprise" Linux distributions today. The number of features and packages that are included to entice software developers, system administrators, or even consumers to consume the product is overwhelming.
However, any given application doesn't need all those bells and whistles, and most OSes are unfortunately not designed to adjust their footprint to the needs of a specific application.
So, the problem isn't that OS capabilities are not needed, just that they are ridiculously packaged, and could in fact be wrapped into software frameworks that hide any division between the application and the systems it runs on.
By the way, that this is exactly why EMC purchased Fastscale last month, as noted by Chuck Hollis, EMC's CTO of global marketing, on the day the acquisition was announced. Simon Crosby, CTO of the data center and cloud division at Citrix, also notes that this change is coming but sees the OS playing a more important transitional role.
This is a critical concept for application developers wondering how cloud computing will affect software architectures. It is also a critical concept for why IT operations professionals need to understand that their roles and responsibilities are changing.
Because of this, I'll be following up with a few posts this week that will expand on this concept and give you much more of a sense of why the operating system, along with most server, network, and storage virtualization is a stop-gap measure as we move to a cloud experience centered on the application user and the developer.
Next on the list is an explanation of why cloud computing drives infrastructure toward homogeneity (at least within a data center) and why that is the bane of server virtualization.
While cloud-computing news this spring has been dominated by the antics of individuals and small groups, a new class of services to support a new class of applications, and today the future of Java, there has been much less excitement about the advances being made in the world of data center hardware to support cloud computing.
This may be, quite possibly, for a very good reason: if you are a consumer of cloud-based resources, the mantra has long been that you can simply deploy or consume your applications/services without any regard to the infrastructure on which they are being hosted. A very cool concept for an application developer, to be sure, but I think it's a mistake to ignore what lies under the hood.
At the very least, the future of hardware ought to touch the inner geek in all of us.
What is happening in data center infrastructure is a complete rethinking of the architectures utilized to deliver online services, from the overall data center architectures all the way down to the very components that serve the "big four" elements of the data center: facilities, servers, storage, and networking.
Here's a quick breakdown of my favorites:
My employer, Cisco Systems, released one of the more visible examples of this, its Unified Computing System product set, which converges compute, network and virtualization into a single integrated infrastructure. I won't go into depth on UCS here--CNET has covered it in depth, as have others. However, the industry is rethinking the practice of buying each component in isolation, assembling systems in a custom fashion each time, and (most importantly) managing them independently. This converged infrastructure is probably the most disruptive change to infrastructure since the commodity x86 server platform itself.
Another company doing amazing things in this space is Rackable Systems, which is taking a wholly different tack. Rather than focusing on the management aspects of the infrastructure, it is focusing on maximum density for minimal energy consumption. Their new platform, the CloudRack C2, is targeted at cloud-computing providers, and is based on lessons learned from some of their largest customers--who in turn are some of the largest cloud providers in the world.
A couple of weeks ago, I spent some time talking to George Skaff, vice president of marketing at Rackable. He talked me through what differentiates C2, and I have to say I was impressed. A tray-based architecture with isolated, temperature controlled, variable speed fans, no power supplies (12V throughout the cabinet, redundant rectifiers for AC-DC conversion) and all wiring on the front of the systems make C2 a truly dense, cloud-ready drop-in server infrastructure.Google added to the fun by revealing its own server architecture. The one thing that stood out here was the placement of a 12V battery right on the motherboard, supporting two processors and two drives per board. However, it is also interesting to note that the entire board is juiced with 12V (not the 12V/5V combination of most commodity boards), and that any power conversion happens on the motherboard.
According to CNET News' Stephen Shankland, "Google's data centers now have reached efficiency levels that the Environmental Protection Agency hopes will be attainable in 2011 using advanced technology." That is extremely cool (no pun intended).
Google server design
(Credit: Stephen Shankland/CNET)Now, why should you care (besides the aforementioned "geekness" factor)? In part, because these are the systems that your future depends on, whether you are a technologist or a business manager. Yeah, the chip sets are familiar and virtualization hides the vagrancies, but this is where your bread and butter lies as you move to data centers architected for the cloud.
Note that there are minor deviances from "traditional" server design here. What I wonder is if (when?) the large cloud vendors will begin to fork their infrastructure designs as they gain more and more control over the data centers that host global IT. When will it become more advantageous to take their custom server design in a direction that supports their custom management and virtualization software--and will that increase the risk to application payloads that should be portable between vendor platforms?
The need for interoperability standards remains great, in part due to this risk. The good news is we have time. I certainly don't think such a fork will happen soon. However, I do believe that it is important that those responsible for IT service level agreements keep track of what their cloud vendors (or their IT internal cloud infrastructure teams) are up to when it comes to hardware.
Whether you agree with me, you have to admit the disruption that cloud computing is having on the data center has made infrastructure somewhat fun to follow again.
If you have an interest in the architectures that may very well come to dominate the world's most sophisticated data centers, you should take some time to check out an article in EETimes, entitled "Server makers get Goooogled."
The article, by Rick Merrit, describes new technologies being introduced by Rackable and other companies that are strongly influenced by Google's custom server designs over the last several years.
We're talking cool stuff here. As the article notes:
... Read More- prev
- 1
- next
