Some things no one seems to be talking about……

Everyone who blogs, tweets (man, I hate that word, but that’s another post), professionally publishes print articles, or has even an audience of 1 in the IT space today is talking about the ‘cloud’.  When you read some of these posts, articles, pontifications, and editorials, you would think that this idea of the ‘cloud’ is going to save mankind from itself!  Our data centers will no longer draw power or require cooling.  Our users will only have to think of something they need and it will instantly be provided for them from this nebulous blob in cyberspace, causing our businesses to achieve new levels of efficiency and causing IT costs to plummet to practically zero.  The ‘cloud’ will instantly cause IT to align to the business because they will no longer be consumed with the worries of managing a data center and finding resources to meet end user demands while being given a bailing wire, bubble gum, and duct tape budget.  If you listen long enough, to enough ‘cloud’ pitches, I’m pretty sure you’ll hear that we’re entering the time of ‘dogs and cats living together, mass hysteria’, and the ‘cloud’ will save us.  Don’t get me wrong.  I am a champion of virtualization, consolidation, and drastically improved management operations.  I’m just not a fan of promises, visions, and theories being promoted as reality.

Sometimes when I read this stuff I feel like I’ve been handed the IT equivalent of Roald Dahl’s Charlie and the Chocolate Factory.  Not the cutesy Mel Stuart/Gene Wilder one, more the creepier Tim Burton/Johnny Depp version, where the Oompa Loompas all look like little clones and seem kind of mean-spirited.  I mean, really, is it just me or does every hardware vendor out there now seem to be talking less about what they can really DO, and more about their ‘vision’.  I’m used to that out of software companies (not that I’ve ever liked it) but seriously I buy hardware for what it can do today, because I need it today, not for what it might be able to do (or not) based on some theoretical vision.  We’re being expected to take a leap of faith into a thing that no two vendors, and in some cases, no two employees from the same vendor, can provide a solid definition and physical road map to get to.  The question becomes will we end up like Charlie Buckett, who wins the entire factory, or Mike TV, who ends up regrettably jumping into a technology he didn’t understand and wasn’t adequately protected from, and thereby end up in the taffy puller (or worse, a ‘cloud’ that ends up spawning a Midwestern tornado storm)?

There are three fundamental, tactical problems I see the ‘cloud’ facing, and no one really seems to want to talk about them.  To the folks I talk to, though, they’re the white elephants in the room.  Over the next few posts I hope to stir up some conversation about them.  As always, I welcome some debate and conversation on these issues.

The first issue I see is:

Bandwidth.  Let’s face it, folks, bandwidth isn’t free.  Bandwidth isn’t cheap, either, especially not the kind of bandwidth we’re talking about to dynamically move compute and storage workloads out of your production data center to some mythical data center in the sky for brief periods of time so your Exchange/Oracle/SAP/<insert your heavy duty app here> users suddenly have a better user experience.  I will readily admit that I don’t have all the details of what this is going to completely take, but it isn’t for a lack of trying.  I’ve asked a lot of questions of a lot of people to show me what this is going to do to WAN resources.  No one really knows.  We do know that performing a migration of several VMs across a LAN can be an unpleasant network event if not designed and planned properly for, so how is this going to play out across the WAN?  Probably not all that well.  There is going to need to be compression and significant optimization of the network deployed.  How?  We don’t really know.  There are a lot of ideas and theories, but if I’m a CIO, I’m not planning the next three years on ideas and theories.  I need to plan on something I can execute with a high degree of certainty that I’ll be successful.  Especially in those environments who could benefit most from the theory of the ‘cloud’ — small to mid-size businesses.  There is no room for error or delay in making these ideas real.  These companies don’t have multi-million dollar budgets that can withstand failed project costs that amount to rounding errors.  There is a responsibility the vendor community needs to step up to here, and in my opinion, they aren’t.

If we don’t tackle the bandwidth issue, adoption of these concepts will be painfully slow, and in many cases will never happen.  We either have to make the bandwidth cheap, or completely mitigate the effect of trying to VMotion over a T1, for instance.  Disaster Recovery and Business Continuity are perfect examples of where this has played out.  How many companies in the mid-market space deploy solid, repeatable, testable DR/BC strategies that involve any kind of network replication?  Very few, despite the fact that they know, some all too painfully, how tapes sitting in a cave or vault somewhere a few miles away from their datacenter are basically useless.  They don’t do it.  The folks I talk to, who readily admit they should be doing something to replicate their data somewhere, repeatedly tell me they can’t afford the bandwidth to do it right.  Technologies like RecoverPoint with its excellent compression algorithms are bringing this to more customers, but still, it isn’t a cheap proposition to replicate large data sets over the pipe.  If we’re going to bring the ‘cloud’ to the masses, this has to change, or the technical details of how it will happen within ‘cloud’ enabling technologies need to come to light, so that there is a solid roadmap to getting there that people can depend on today.

Remember, we aren’t talking about Web 2.0 (another meaningless term that somehow became a word) technologies that are working through a browser and are designed to work across mid to high latency Internet links.  We’re talking about the need for links that allow us to dynamically move workloads based on the needs of those workloads and end user demand placed upon them.  Also remember, that when you transfer the workload, you’re going to have to have the data there, too, and this is the same data that you already aren’t replicating due to bandwidth constraints/costs. 

This is not a light or an easy point to overcome.  Will someone please tell me how this is going to be addressed, especially in the context of the SMB market, who stands to benefit the most from these dynamic shifts of peak workloads?  When I ask these questions to vendors, it’s all bull frogs and crickets.  Put another way, I feel like the Oompa Loompas are about to come out from the wood work with all the dancing and singing that gets done.  If I’m wrong, and someone has developed a good answer, but I just missed it (which is entirely possible), will someone please point me in the right direction?

Next up…….Security.