VMware, Hyper-V… XenServer? The choice is yours
VMware, Hyper-V ... XenServer? When it comes to virtualisation, these are the three most frequently cited options. And this would have been OK in the days before cloud, when virtualisation was “just” something for the boys and girls down in the sysadmin branch of the IT department cared about. Now we do have cloud, and private cloud at that, everybody reckons they have a stake in deciding what’s best.
So... what is best? Wrong question. First thing to ask is why VMware, HyperV and XenServer?
Formed in 1998, and launching its first desktop virtualisation product – VMware Workstation – a year later in 1999, VMware is the old guard and the standard everyone is familiar with. VMware moved into the server market in 2001 with GSX (a hosted product, which later became VMware server) and ESX, the enterprise VMware product.
To say it has ballooned from there would be something of an understatement.
After a couple of years came VMware ESX and the two biggest strings to its bow have come to set what we expect as a bare minimum from an enterprise hypervisor: the central management system (vCenter server) and the live virtual machine migration (vMotion). Then, 64-bit support, live storage migration, advanced distributed networking, automated workload management and high availability soon followed, raising the standard even higher.
The Xen Project was initially developed and launched by Cambridge University in 2003 with Xen bought by Citrix in 2007 as a commercial enterprise – although the free, open-source version still exists and is well-supported by Citrix – and has followed swiftly after VMware’s product with the increased feature set; core components like 64-bit support and live VM migration (XenMotion, in Citrix parlance) have kept XenServer competitive with VMware in terms of features.
Microsoft is the latecomer, and released the first enterprise Windows hypervisor – Hyper-V, known as Project Viridian throughout development – to a relatively muted reception in 2008.
The first iteration of Hyper-V was treated with a certain degree of skepticism; the idea of operating an always-on enterprise hypervisor, providing the underpinnings of dozens (or hundreds, even thousands) of other VMs all the while running on a fundamentally Windows platform was met with ridicule and snobbery by the established Linux-kernel-based hypervisor market.
Hyper-V versions 2 and 3 turned out to be rather excellent and are now following the same path as XenServer, delivering equivalent features to VMware’s headline acts in their own inimitable style.
Before you compare this trio you need to work out the operating environment to ensure a level playing field.
Configuration maximums are, much like benchmarks for processors or maximum IOPs figures for storage arrays, largely theoretical. Yes, you can achieve 400,000 IOPs if you load the array entirely with SSDs, but that’s not exactly practical.
The same is true of configuration maximums on hypervisors to a certain extent – I’m yet to meet anyone who has actually deployed a single 4TB VM – but they are a good indicator of the relative size and scale of each solution.
Additionally, you’d do well to learn the configuration maximums if you’re planning on sitting a professional certification like the VMware VCP, as they always come up. Until recently, Citrix would actually ask what year Citrix (the company) was formed on every single exam – if you’re interested, it was 1989 – and the name is a portmanteau of Citrus and UNIX. Obviously.
Here’s some of the more useful maximums when comparing hypervisor platforms:
Configuration Maximum | ESXi/vSphere 6.0 | XenServer 6.5 | Hyper-V 2012 (v3) |
---|---|---|---|
Cluster nodes | 64 | 16 | 64 |
VMs per cluster | 8,000 | 4,096 | 8,000 |
CPUs per host | 480 | 160 | 320 |
RAM per host | 12TB | 1TB | 4TB |
VMs per host | 1,024 | 500 | 1,024 |
vCPUs per VM | 128 | 16 | 64 |
RAM per VM min | 4TB | 192GB | 1TB |
RAM per VM max | 64TB | 2TB | 64TB |
So, while the configuration maximums of Citrix XenServer are lagging slightly behind that of VMware’s offering, it’s reassuring to see that Microsoft have all but caught up in this regard.
It’s also worth mentioning that some of the maximums – like virtual disk size – are limited to 2TB in older versions of vSphere and Hyper-V; it’s only with the latest versions and new virtual file systems that these limits have leapt to 64TB.
Additionally, the XenServer configuration limits are based on the commercial product and Windows guest VMs; if you were running Linux VMs on a custom implementation of the Xen Project (perhaps with a Cloud Stack control framework) you could likely push these much further.
As I said, it’s worth taking them with a pinch of salt, but they do serve as a good thumb-in-the-air indicator of capacity and performance.
From the earliest days of simple virtualisation, the dramatic development and availability of features in the enterprise hypervisor turned them from a basic P2V tool for reducing hardware into a powerful arsenal of abilities and tools that a physical server platform couldn’t even begin to dream of.
Here’s a run-through of some of the best and most important features, by platform, starting with the forerunner – VMware vSphere – and their XenServer/Hyper-V equivalents:
vCenter: The centralised management solution for a VMware virtual data centre. This is the beating heart of your virtual environment and without vCenter, none of the amazing features would be possible. Has recently transitioned from a Windows VM with a SQL/SQL Express database requirement to a dedicated appliance with built-in custom database.
vMotion: Live VM migration between hosts without downtime. Requires shared storage and networking between hosts, plus a low-latency link, but in recent iterations can be carried out long-distance.
Storage vMotion: Similar to vMotion, but used to migrate running virtual machines between different data stores for live storage migration.
High Availability: Utilising heartbeat connectivity checks between hosts (and additional, individual VM monitoring if required) to ensure virtual machines stay up during any eventuality. If a host is lost through hardware, network or power failure, virtual machines will be automatically brought up on other hosts with spare capacity.
Fault Tolerance: During a regular High Availability failover, virtual machine uptime will be lost for the time taken for the host loss to be detected, plus the time taken for virtual machines to restart. Fault Tolerance maintains a second, standby copy of a virtual machine on a second host running milliseconds behind the original, ready to go immediately in the event of primary host loss.
Distributed Resource Scheduling: An automated workload management solution that distributes virtual machines between hosts according to resource requirements, to ensure the most efficient use of the virtualised platform. Can be fully automated or notify administrators of virtual machine moves that will better distribute the workload.
Distributed Power Management: An additional automated workload management solution that makes use of integrated lights-out management (HP iLO, Dell iDRAC, IBM IMM) on host servers to power hardware off at times of low-load to save on power usage, then switch them back on when the workload demands.
Hardware Hot-Add: Allows for the on-the-fly addition of vCPU, RAM, hard disks or other peripherals to be powered on and running virtual machines without disruption. Requires a supported guest operating system; hot-add is supported in Windows Server Enterprise versions, but not Standard.
vSphere Replication: Allows replication of virtual machine clusters between multiple data centre environments, without relying on an underpinning storage array replication. Can be used in conjunction with Site Recovery Manager for a full disaster recovery solution.
vShield: A relatively new security suite within VMware that – when combined with a compatible storage vendor – allows you to offload anti-virus and other security software functions to a dedicated control appliance, reducing installed software and performance overheads on protected VMs.
So taking all that into consideration, which hypervisor comes out on top in our comparison? That’s a very sticky question and one you may struggle to find a definitive answer to.
There’s no doubt if you’re chasing after the latest features or require the biggest configuration maximums, then VMware should absolutely be at the top of your list.
It generally offers the best combination of companion products too, like networking solution VMware NSX or disparate storage aggregator the VMware Virtual SAN. Its solution doesn’t come cheap, though, and there’s a high degree of complexity which might serve as an additional barrier to entry.
If you know Windows, as most sysadmins do, then working with Hyper-V is like putting on a comfy old pair of slippers. It’s easy to use and now is (just about) as powerful as the competition, yet is in the middle-of-the-road in terms of costs.
And XenServer? The configuration maximums may be smaller but it’s a gutsy little performer that is operationally as effective as Hyper-V and vSphere, yet could save you an absolute fortune on licensing costs.
The choice is yours.