How Virtual Machines Solved Moore’s Law

By: Amir Sharif 02.25.2019
How Virtual Machines Solved Moore’s Law

Moore’s law is the observation that the number of transistors in a dense integrated circuit doubles approximately every two years. This trend was first spotted in 1965, by Intel co-founder Gordon Moore. Moore’s Law states that a CPU will reliably get bigger and more powerful at a consistent rate, creating a problem for enterprise IT centers.

When a fleet of servers that are becoming progressively more powerful remains bound to one application per OS per server model, they become less efficient and more troublesome, especially at scale. The standard practice is to bind one application per OS to simplify troubleshooting and minimize conflicts. However, as Moore’s law tells us, this will lead to massive efficiency issues that get progressively worse.

Imagine having an airline where every few years you’re adding bigger jet engines and more seats to your plane so your capacity is increasing. But because you have a certain process, you cannot increase the number of passengers per plane. So despite the planes getting bigger, faster, and more capable, you’re transporting the same number of passengers back and forth, and thus your relative efficiency is coming down. So you are not capitalizing on the power of the server in real terms.

About 15 years ago, servers were getting so powerful, the utilization rate was in single digits. 

To solve this problem, VMware popularized the notion of the hypervisor. The hypervisor is a very specialized OS that brokers access to a very specific type of application. The hypervisor is a process that creates and runs virtual machines (VMs). VMs are neat packages that allow you to maintain operational consistency.  A hypervisor allows one host computer to support multiple guest VMs by virtually sharing its resources, like memory and processing.

That application is hardware-created software; emulated hardware running on physical hardware. You’re still preserving the model of one application per OS per server. But using you can pack a lot more resources per physical server. The NIC has one or more IP addresses, and there is now a virtual NIC, one or more of them per our virtual machine. So the number of IP addresses is also increasing as the number of VMs increases too.

The VM becomes a chunk of information that’s being executed by the hypervisor. Because it’s software, you can now store it in a disk, you can reboot it much faster, you and can move it around. Because of these properties, it becomes more ephemeral than the physical server underneath.

In summary: Though earlier in the century, your server’s power was woefully underutilized due to Moore’s Law, we live in better times.  With hypervisors, one host computer can support multiple guest virtual machines. Thanks to a simple process of emulation, you can keep the model of one application per OS per server, but you can pack a lot more resources on to physical servers.

For Moore’s Law, the jig is up. In 2016, MIT’s Technology Review declared, “Moore’s Law is dead.” In January 2018, the Register said the same thing and issued a “death notice” for Moore’s Law.

Moore’s Law may be dying.  Its legacy is the massively powerful CPU.  Long live virtual machines!

Recent Posts Of Ranchers and iPads: How British Columbia Replaced Paperwork with OpenShift and Aporeto The End of the Monolithic Application Simple by design; Automating per-namespace isolation with Aporeto and OpenShift