Back to School – Infrastructure 101 – Part 2: Virtualization
scale last edited by scottalanmiller
I covered SAN technology in my last Infrastructure 101 post, so for today I’m going to cover server virtualization and maybe delve into containers and cloud.
Server virtualization as we know it now is based on hypervisor technology. A hypervisor is an operating system that allows sharing of physical computing resources such as networking, CPU, RAM, and storage among multiple virtual machines (sometimes called virtual servers). Virtual machines replaced traditional physical servers that each had their own physical chassis with storage, RAM, networking, and CPU. To understand the importance of hypervisors, let’s look at a bit of history.
Early on, computing was primarily done on mainframes, which were monolithic machines designed to provide all of the computing necessary for an organization. They were designed to share resources among various parallel processes to accommodate multiple users. As computing needs grew, organization began to move away from the monolithic architecture of the mainframe to hosting multiple physical servers that were less expensive and that would run one or more applications for multiple users. Physical servers could range in size and capacity from very large, rivaling mainframes, down to very small, resembling personal computers.
While mainframes never disappeared completely, the flexibility in cost and capacity of physical servers made them an infrastructure of choice across all industries. Unfortunately, as computing needs continued to grow, organizations began needing more and more servers, and more administrators to manage the servers. The size of server rooms, along with the power and cooling needs were honestly becoming ridiculous.
There were a number of technologies that emerged resembling what we now call server virtualization that allowed the compute and storage resources of a single physical box to be divided among different virtualized servers, but those never became the mainstream. Virtualization didn’t really take off until hypervisor technology for the x86 platform came around, which happened at the same time as other platforms were declining in the server market.
Initially, virtualization was not adopted for production servers but instead was used extensively for testing and development because it lacked some of the performance and stability needed for production servers. The widespread use for test and dev eventually led to improvements that made administrators confident with its use on production servers. The combination of performance improvements along with clustering to provide high availability for virtual machines open the door for widespread adoption for production servers.
The transition to virtualization was dramatic, reducing server rooms that once housed dozens and dozens of server racks to only a handful of server racks for the host servers and storage on which all of the same workloads ran. It is now difficult to find an IT shop that is still using physical servers as their primary infrastructure.
While there were many hypervisors battling to become the de facto solution, a number of hypervisors were adopted including Xen and KVM (both open source), Hyper-V, and VMware ESX/ESXi which took the lion’s share of the market. Those hypervisors or their derivatives continue to battle for marketshare today, after more than a decade. Cloud platforms have risen, built over each of these hypervisors, adding to the mystery of whether a de facto hypervisor will emerge. But maybe it no longer matters.
Virtualization has now become a commodity technology. It may not seem so to VMware customers who are still weighing various licensing options, but server virtualization is pretty well baked and the innovations have shifted to hyperconvergence, cloud, and container technologies. The differences between hypervisors are few enough that the buying decisions are often based more on price and support than technology at this point.
This commoditization of server virtualization does not necessarily indicate any kind of decline in virtualization anytime soon, but rather a shift in thinking from traditional virtualization architectures. While cloud is driving innovation in multi-tenancy and self-service, hyperconvergence is fueling innovation in how hardware and storage can be designed and used more efficiently by virtual machines (as per my previous post about storage technologies).
IT departments are beginning to wonder if the baggage of training and management infrastructures for server virtualization are still a requirement or if, as a commodity, server virtualization should no longer be so complex. Is being a virtualization expert still a badge of honor or is it now a default expectation for IT administrators? And with hyperconvergence and cloud technologies simplifying virtual machine management, what level of expertise is really still required?
I think the main take away from the commoditization of server virtualization is that as you move to hyperconvergence and cloud platforms, you shouldn’t need to know what the underlying hypervisor is, nor should you care, and you definitely shouldn’t have to worry about licensing it separately. They say you don’t understand something unless you can explain it to a 5 year old. It is time for server virtualization to be easy enough that a 5 year old can provision virtual machines instead of requiring a full time, certified virtualization expert. Or maybe even a 4 year old.
scottalanmiller last edited by
Mod note: Added "Virtualization" to the title as that was needed to not be confusing.
scale last edited by