This will sound like a reincarnation of Rich’s ‘think-before-using-technology’ post, but I must admit that in the short life of this blog I’ve became addicted to them, and you know, bad artists copy, real artists steal, so here I go like a bad artist.
No doubt this will be the year of the flat data center. We heard a bit of it at the end of 2010, but we are in the second week of 2011 and most IT publications already have devoted some pages to the topic of the flat data center. You don’t need to turn on your brand new crystal ball (flat crystal ball, of course) to know that flat data centers will be the flavor of this coming summer.
Network World touched on this in their predictions article. If nothings goes wrong, this year we expect to see TRILL and Shortest Path Bridging (SPB) (802.1AQ) ratified in their respective certification bodies. It is a pity that the industry couldn’t come to an agreement again, and leave it to the market to choose which one is better fitted to the enterprise. With big powerhouses on each side the choice won’t be easy.
In little time you will see vendors choose sides and you will have to make your choice for one or another.
Or maybe not… Like Rich has taught us in his posts, technology makes no sense at all if it doesn’t add value to our business. In my spare time I test fancy protocols, obscure OSes, brand new services and shiny new devices. Enterasys expects a different attitude from me while I am on working hours. And following this attitude I still have to see an architectural solution that adds value to a business.
I’ve tried everything in the last fifteen years: L2 LAN emulation over ATM, spanning tree, L3 campus networks, Ethernet, token ring or FDDI, Enterasys SecureFast, Flat designs based on Ethernet L2 multihoming without spanning tree and multidevice LAG. And none of them added a penny to the value of the business.
What was adding value was the ability of the providers and integrators to mix the features at hand to benefit IT work flows and operations.
Raw architectural designs have some issues. First, you cannot mix them. When my customers decided that L2 LANE was a dead end, they had to trash all their infrastructure to embrace Ethernet and spanning tree, not a change taken lightly. Second, you never know what to expect from new architectures until you use them, and you must be very disciplined to take full advantage of the new architectures. Badly managed spanning tree can be the worst of all nightmares, but so can chassis clustering solutions. In my career, I spent many a night awake diagnosing STP as I did chassis clusters that aimed to remove STP, LANE networks or the Gigabit Ethernet technology that replaced them. I might remind readers that LANE and SecureFast are closer to TRILL and SPB than you may think.
This is without even mentioning the influx of pre-standard solutions that will be offered before standards are ratified. I’d like to speak more about this, but I’ll leave that for a later post.
So, before embracing the new architectures like a shiny new iPad, think twice what will be the value add to the business. For instance, recovery time: I see SLA contracts that charges for downtime longer than the 30 seconds of STP. You can make it null (but, remember, that is to be shown in real environments) with the new architectures, but what for? Actually, 30 seconds of STP reconvergence is peanuts compared to the downtime caused by other means, sometimes due to the complexity added by the protocols aimed to replace STP.
What do you think? Which standards are you most looking forward to?