The Beginning of the End of IT as We Know It

Arthur Cole

Continuing with our look at year-end reviews and predictions, I'm struck by the way many of these reports focus on individual technologies but not on the data center in general.

If your focus is on, say, the future of servers or virtualization in 2010, there's no shortage of opinion out there. But as an infrastructure guy, I'm more interested in how everything stands to come together -- not just in the next year but over the next decade.

To get that insight, we sometimes have to dig a little deeper into what people are saying. Take cable giant Media General, for instance. Mike Miller, director of information security, tells IDG that the company is pursuing a broad virtualization and consolidation strategy in 2010, mostly to make up for lost ground during budget-challenged 2009. To lay the groundwork, the company has been stocking up on quad- and six-core, dual-socket machines -- the kind with enough horsepower to host upwards of 50 VMs. Similar stories are heard from Qualcomm and Scottrade.

What that tells me is that virtualization and cloud computing are likely to reverse last decade's trend of multiple, smaller servers and send us back to the heavy artillery that came before, albeit with better performance and lower power consumption. At the least, that calls into question the strategy of Intel and others to develop the "microserver" and other technologies designed to put processing power into smaller and smaller packages.

The idea there is to develop systems for smaller organizations, so they can stuff all of their core IT infrastructure in a broom closet rather than dedicated server and storage rooms. But that notion is looking increasingly iffy in an age where SMBs will likely turn to the cloud for their IT needs, employing a mix of application, platform and infrastructure services rather than wrestle with their own internal hardware, no matter how small it gets.

That leads us to probably the biggest change facing the IT industry: the end of on-premises IT infrastructure, save for the largest organizations, and the rise of outside services. As demonstrations at Cloud Expo and other showcases are making plain, organizations of all sizes could obtain all of the capabilities they have now through PaaS tools like 3Tera's AppLogic or by accessing cloud storage via a simple Layer 2 switch rather than a full-blown SAN.

And if that is the case, then we're not looking at just bigger and badder hardware, but bigger and badder data centers. Why should all the benefits of consolidation end at the data center wall? As companies like Unitiv are pointing out, data center consolidation offers all the benefits of improved manageability, scalability, availability and the like that hardware consolidation provides, but on a larger scale because we're talking about shutting down entire facilities rather the little boxes.

So in the end, what we're likely to see within 10 years' time are massive, regional data centers populated with heavy server equipment and racks of storage tied together with high-speed networking (InfiniBand, anyone?), all doling out resources to anyone with a bank balance. Some of these facilities will come from the top-tier enterprises, which will still use them for internal purposes, perhaps leasing off any unused capacity. But others will be built by dedicated data center providers, making IT available over the Internet much the same way electric companies deliver their product over the grid.

And that's when we'll say goodbye to IT as another cost center in the larger business picture and welcome it as a distinct industry in its own right.

Add Comment      Leave a comment on this blog post
Jan 8, 2010 7:26 AM Ken Cameron Ken Cameron  says:

I believe there are "pieces" of truth in your predictions.Exactly how they will come true is yet to be seen.Part of me says you go way too far "out there", yet another part of me says you haven't taken it far enough.Technology is finally evolving to a level where mega-center grids are becoming possible, but we are not quite there.I can easily shift workload dynamically within a center.Shifting workload dynamically between centers still needs some work.One hurdle is that, today (at least in enterprises), many applications can be and are spread across multiple platforms (mainframe, UNIX, and Wintel) and multiple storage venues (SANs, NAS, DAS, etc.).How do we move such complexity into the cloud, WITHOUT a total re-architect &rewrite?I see the cloud market dividing.We will have a consumer &SMB market that will quickly move down the cloud path, including, possibly, mega-centers.However, the enterprises will stay with Private Clouds, and until some "Enterprise-Class" vendors start providing Private Cloud services, the enterprise market will stay in-house, except for pure-play external/private cloud (which is called outsourcing).Gradually, enterprises will go hybrid, adding external resources to their private clouds (capacity expansion and technology refresh), and this concept will grow linearly at first, then exponentially until the majority of enterprise services are external (10-20 years from now).This will partially depend on the RIGHT vendors coming to market with such offerings.A possibility is that the largest enterprise data centers will just "join-the-grid", even selling excess capacity back to the cloud.Re-reading your article, I almost get the feeling you are saying it is time to go back to the mainframe ("/> ), which I do believe will be an increasing trend.With the z/10 and coming z/11 processors, I can run hundreds and even thousands of virtual instances on ONE server, which is, by the way, the most reliable, most scalable, most serviceable, most secure, most "green" server on the market today.Now, you can even run z/OpenSolaris, and z/Windows.IBM has even announced z/LINUX ONLY mainframes.All of the potential cloud providers should take note that IBM is just now beginning to hint at the mainframes position in the Cloud market.They need not worry, though, because the mainframe is dead!I read about it 2 decades ago.Right!!!Above, I mentioned "Enterprise-Class" providers.A major hurdle for the coming external Cloud market is going to be the track record of some of today's big name providers.The 2009 ouages at Amazon, Google, Microsoft, Rackspace, the providers at the Foster Center in Seattle, are very well known.Yet, these players are all touting their Top-Tier data centers, and guaranteed service levels, blah-blah-blah.Consumers and most SMBs probably didn't understand the implications when Microsoft announced that the December BING outage was due to a mid-week, mid-day change that did not have a viable back-out plan.Ask any enterprise IT person what that tells them about the maturity of Microsoft's operations.The "reported" root cause of several of the facility outages told me that there are gaping holes in their N+1 designs, not to mention, their backup & Reply

Jan 8, 2010 7:26 AM Ken Cameron Ken Cameron  says:
DR plans.It will be a long while before large enterprises put all of their jewels in the hands of these Enterprise-Class "Wanabe" vendors.IBM, CSC, ACS/Xerox, Perot/Dell, UNISYS might be okay.Note that I leave HP/EDS off the list after their recent debacle (mainframe outage) with RBS, apparently caused by an HP executive management decision to layoff the mainframe staff (could only happen at an Open Systems vendor).Arthur:no matter how this goes, it is going to be an exciting decade.Instead of seeing this as the end of the company data center, I see this as a declaration of independence by the infrastructure side of IT, AND, the potential merger of everyone's data centers.The glass is "Half-Full".

Ken Cameron

Jan 25, 2010 6:06 AM Joe Gleinser Joe Gleinser  says:

I agree with your projections IF bandwidth prices decrease and quality increases at a much faster rate. True, low quality bandwidth is quite cheap. When security, QoS and latency are factored bandwidth is not nearly cheap enough to justify a complete cloud/PaaS solution for most businesses. Egnyte's Local Cloud and VMWare's vCloud Express are much more likely to be the standard in the near future.


Post a comment





(Maximum characters: 1200). You have 1200 characters left.




Subscribe to our Newsletters

Sign up now and get the best business technology insights direct to your inbox.