There is a lot of froth being whipped up over cloud computing and everyone is digging out their old Platform War playbooks. This time, they say to themselves, we'll do it right. But this platform shift is a bit different from the last one, and the lock-in that bites you might not be the one you expect.

Companies are right to worry about vendors locking them in with incompatible software. All of the same old tricks are there waiting to be pulled. (My favorite is Amazon charging less to put data in than to take it out.) We as customers don't want the vendors to compete on tricks and FUD but rather features, quality and price. We want open standards for data and APIs from Cloud to Cloud.

But we'll still be locked in because of network latency.

Imagine that you have some virtual servers and data in Cloud A. But Cloud B has a neat new service that can do something cool with your data, like detecting fraudulent activity. You start testing it by sending data from Cloud A to Cloud B's API... which is horribly slow, error-prone and expensive.

A packet of data sent between Clouds has to pass through two sets of firewalls and the regular internet. The network distance between Clouds is much greater than within, where everything runs on souped-up private networks usually inside the same building. And of course you pay twice for the packet. This makes a lot of kinds of mix-and-match between A and B expensive and impractical.

Latency is the perfect kind of lock-in because no one can be blamed for creating it and every vendor gains by ignoring it. Your data is here and the other Cloud is over there. In between is a messy, slow road with toll booths on either side. That's just the way it is. Better to keep your data inside the bubble where ping times are zero and transfer is free.

In the meantime the vendors get to have a chest-beating party over standards and openness. But really they can't lose. If they pull off API lock-in they win. If not they still have latency lock-in and can cut deals when clients want to move to (move to, not interoperate with) different Clouds. So far big IT sites are only dipping their toes. Their first apprehensions are always about the data but the answers are always about open standards and how insecure doing it yourself is.

It all comes down to your data and the cost of moving it around. In a way it's like banks. They take care of the headache of storing and transferring money. No modern business has to keep a cash vault in the basement to make payroll. But in return the banks take a cut of every transaction. Behind all the cloud talk is the wet dream of tech vendors: taking a cut of every computation. There will always be holdouts but I don't think anyone can stop it. Right now we should be thinking really hard about how to strike this deal.

What's a CTO to do?

How about we demand "Cloud Peering"? It's an analogy to network peering: two companies that own lots of internet cables agree to build a special high-speed connection between them, route each other's data, and write off the cost. Kind of like how the Post Office will deliver a letter with a UK stamp and vice-versa. Peering (and for that matter, mail) is much more complex than that but the point is that complexity is hidden.

It will never be perfect -- you can't wish away congestion and the speed of light. We will have to learn to build systems that are more asynchonous and resilient.  But you as a cloud user can demand simple, efficient data interchange at a fixed price and the vendors can make it happen.

A lot of smaller startups are fully in the cloud but their Spidey senses are tingling. They should try to stay nimble and encourage vendors to agree on compatible software. Mosso has the right idea by supporting an Amazon-compatible API and adding features around it. At Archivd we're looking at running AWS and Mosso in parallel.

Larger CTOs can help by asking now about latency lock-in and cloud peering. Without pressure on the vendors it'll never happen. Whether they know it or not, right now CTOs have the upper hand. Once they hand over the data, their leverage is gone.

Leave a Reply