Steve Bondy

I’ve been watching the cloud computing discussion grow from a distance for a while now, and I think it’s time to jump into the discussion with both feet.

There’s an old expression about what happens if you don’t remember the past, so I’d like to suggest that the cloud computing initiative is in some ways the completion of a full cycle of computing evolution. Step into my time machine, as we revisit the 1960’s and 1970’s.

In those times, compute resources were centered on mainframes - expensive devices offered by a few vendors such as IBM, CDC, and Univac. Due to the cost of these computer resources, many companies could not afford their own data centers, and instead purchased “time” on a system owned and maintained by another organization. The bill to the customer was usually divided into different types of “time” - connection time, CPU time, peripheral time, etc. I have memories of my CS classes where when logging out of the system I was shown how much my account was debited for the time used. As an aside, being a successful CS student didn’t just mean getting your projects done, but getting them done without exceeding the credit in your account.

This model held sway for many years, and one of the significant developments in the data center was the release of Digital Equipment Corporation’s VAX running the VMS operating system. Now user tasks were segregated into virtual machines, isolating them from each other and prioritizing their execution.

But the centralized model changed into the desktop model, as more and more organizations found that relatively inexpensive computer resources could be made directly available to users, without an expensive data center. This model them morphed into the client-server model as fewer, more powerful servers began to take over compute intensive tasks such as databases, and e-mail.

And then a funny thing happened. As the servers got more powerful, their ability to process data exceeded the demands placed on them by the applications. The unused processor time was wasteful, so some enterprising vendors took a giant step - backward - to virtualized machines.

But that was only the first retrograde motion. The second one is coming in the form of cloud computing. Once again, compute resources will go into amorphous data centers, with a focus on sharing those resources across a wide user base. Once again, there will be little in the way of physical resource identity. Walking into a data center and asking to see the server running a particular application will be a difficult question to answer, as smart virtualization techniques shift VMs between computer resources as needed based on workloads. Billing to the customer will likely return to something like the old model, listing the number of VMs, and charging based on their impact on the overall computing fabric, with other billing categories for storage and bandwidth.

I’m not suggesting any of this as a bad step, but it is interesting to note that we seem to have come full circle in many ways. And if that is true, what will happen next? Will we see the “cloud” give way once again to distributed resources? I don’t know - but I do know it will be interesting!