Virtualize Everything

Be on the lookout! Anything in information technology (IT) that hasn’t been virtualized already is a target for abstraction. This series certainly assumes host virtualization, and has already discussed virtualization of the other key computing elements: storage and networking.It doesn’t stop there, though! Virtualization is also possible at the highest levels as software-defined datacenters (SDDC).

SDDC is simultaneously a natural extension of proven techniques, and also an occasionally blunt marketing ploy to focus attention away from the promise by Oracle and a few other integrated vendors to sell highly-tuned end-to-end, top-to-bottom Exa- or similarly-branded “solutions”. Along with Oracle, a handful of other vendors, including HP and IBM, catalogue what they offer as turn-key products. Buy these solutions, and you won’t have to worry about their parts again, because the vendor pledges they all work together smoothly. More precisely, if a problem does turn up, that one vendor will take care of it, no matter where in the stack it fits.

SDDC takes nearly the opposite approach. Instead of a highly-integrated, tuned, and even certified configuration, SDDC treats everything having to do with computing as a commodity fit for almost arbitrary combination with other standard virtualized pieces. What’s the appeal in that? At the executive level, it gives “enterprises the ability to import existing images … control their own network security …” and customize and automate to manage scaling and integration challenges.

The power of this kind of “commoditization” has far-reaching consequences, several of which we’ll explore in future postings. Here’s one compelling instance: suppose a new enterprise is undertaken–financing arranged, plans approved, and so on. The business needs a datacenter. Old business models would take months to build a datacenter, or weeks to negotiate a colo configuration, or perhaps, in the best case, days of intense effort to situate and connect a datacenter-in-a-box or set up resources in the cloud. With SDDC, though, all the baseline computing a business needs can, in principle, be spun up in hours or minutes! Entrepreneurs should be putting their ideas into practice, not waiting on racks to be assembled and cooling systems plumbed, and SDDC will enable these dramatic leaps in IT agility.

At the same time as SDDC woos top-level decision-makers, virtualization and closely-related abstractions are also stepping into the lowest levels of computing. Microcode and field-programmable gate arrays (FPGA) are a couple of now-commonplace technologies for making central processing units (CPUs) more flexible. Now Intel announces that its photonic research has advanced almost to the point of breaking free from the conventional board and rack. I like Rich Miller’s coverage, at the previous hyperlink, because it emphasizes the architectural significance of photonics. Certainly, raw performance boosts are welcome; more interesting, though, is the way photonic advance commoditizes at the lowest computing levels, and “enables independent upgrading of compute, network and storage subsystems”, as Intel’s chief technology officer Justin Rattner is quoted. This means, for example, that a central processing units (CPUs) might be upgraded in-place–without having to refresh a whole server or even motherboard.

You do not have to keep up with all these innovations; in fact, you cannot. Not even billion-dollar operations like Google or Amazon are tackling every virtualization experiment simultaneously. You do owe it to yourself, however, to cultivate a virtualization mentality. If your operations have any “pain points” subject to tedium, human error, unreliability, clumsiness, or out-of-control costs, look for a way to virtualize the problem into submission. Virtualization doesn’t heal broken parts or reduce vendors’ list prices, of course–what it has done and will do in the future is to make operations more predictable, reproducible, scalable, nimble, and therefore manageable. Once virtualization provides its benefit of more consistent, predictable and even “agile” operation, you have the chance to research how to fix the breakdowns within the virtualized parts.