Let’s go wild here. Say you’re uncertain whether to keep your brain as is. You think a certain Harry Potter-type surgeon could create a better version. But you're not sure. You’re also afraid this surgeon will scotch up your brain while he fiddles on improvements. So you have the surgeon construct cabinets in your skull - these cabinets are on the periphery of your brain - where he does all his work. If his virtual brains are better than the brain you have now, the surgeon replaces your brain with his creations. If they’re not, the surgeon continues producing virtual brains in his cabinets until you’re satisfied with his results.
Those cabinets are called virtual machines. The layer that overrides and organizes these cabinets as well as giving the surgeon more room to work in, is called the hypervisor.
In the computer world, we have the hardware which is the equivalent of your body, and the software, the equivalent of your brain, that drives the body. Now, say you want to improve some existing software but are afraid that tinkering on it could irreversibly destroy the original system. Computer engineers solved that problem by building one or more virtual machines—or virtual cabinets—(like mini labs) where they tinker on their prototypes, called instances, while the original stays intact.
At one time, this software tool was called the “supervisor”. It’s the additional digital layer that connects each of your virtual machines (VMs), supervises the work being done in the VMs, and separates each VM from the other. In this way, your instances are organized and your VMs are rendered coffin-tight to outside interference, protecting your instances, or innovations.
You’ve got two types of hypervisors: Those that sprint side-by-side the VMs and those that shimmy on top. In either case, the hypervisor serves as an “alley” for storing additional work information.
Amazon’s Nitro Hypervisor
Nine years ago, Amazon Web Services (AWS) noticed that very soon software developers would have a problem. The hypervisor system was wasteful; they consumed too much RAM, they yielded inconsistent results, and their security would be challenged with the accelerating bombarding software.
“What we decided to do,” Anthony Liguori, principal software engineer of Amazon and one of the key people who planned and executed the venture told me, “was to completely rethink and reimagine the way things were traditionally done.”
The VMs and hypervisors are software. So, too, all the elements—input/output (I/O) functionalities—are integral to these systems. What AWS did was tweeze out each of these I/Os bit by bit and integrate them into their dedicated hardware Nitro architecture, using a novel silicon produced by Israeli startup Annapurna Labs.
Today, all AWS virtualization happens in hardware instead of software, shaving management software costs and reducing jitter to microseconds.
Since 2017, more companies have emulated AWS and likewise migrated most of their virtualization functionalities to dedicated hardware, in some cases rendering the hypervisor unnecessary. This means all virtualization could now be done from their hardware tech stack without need of a hypervisor.
Virtual machines are for deploying virtualization models, where you can build and rebuild instances at your pleasure while protecting the original OS. The hypervisor operates and organizes these VMs and stores additional work information.
In the last few years, AWS developed its revolutionary Nitro virtualization system, where software VMs and hypervisors were transmogrified into dedicated hardware form. In this way, working on instances becomes cheaper, faster and more secure. Innovations also unfurl faster, since both VM and hypervisor layers are eliminated. More vendors, like VMWare, Microsoft and Citrix, emulated Amazon and introduced their own so-called bare metal hypervisors, too. These hypervisors are called Type 1.
Meanwhile, Google Cloud uses the security-hardened Kernel-based Virtual Machine (KVM) hypervisor. KVM is an open source virtualization technology built into Linux, basically turning Linux into a system with both hypervisor and virtual machines (VMs). Although it’s Type 2 (since it runs on top of an OS), it has all the capacities of a Type 1.
Leah Zitter, PhD, has a Masters in Philosophy, Epistemology and Logic and a PhD in Research Psychology.