UPDATE: This information is based on older AOS 4.6 version , for comparison between AOS 5.5 and Vsphere, plese read updated version.
For most implementations requiring basic hypervisor functionality Nutanix KVM-based Acropolis Hypervisor (AHV) is a good choice, even with some functional benefits over Vmware vSphere, cost wise AHV is hard to beat as it comes for free with Nutanix system. Please visit Josh Odgers’ blog for a good description of benefits using AHV.
However there is less information available on cases where Vmware hypervisor might be better fit or the only option. Like always there are fringe cases where a solution isn’t optimal or doesn’t fit at all, that is true also for AHV.
Nutanix has never aimed to have feature parity with Vmware, but is actively developing new features to narrow the gap. Following limitations/differences are based on AHV found in 4.6 version, these might be resolved in future releases. If you think that I am wrong, please leave a comment and I am more than happy to correct any mistakes.
List of features where differences can be found. Not in any particular order and not an all-inclusive list. I am sure that there are more differences, these are the ones I’ve encountered while talking about Nutanix with potential customers.
- GPU support
- Affinity and Antiaffinity rules
- Metro Availability
- Guest Customization
- Open vSwitch Management
- Role Based Access Control (RBAC) (future post)
- VM Memory Utilization & Memory Sharing (future post)
- Multi-VM Operations (future post)
- Automated VM Migration (future post)
In order to use Nutanix native snapshots with backup applications, the backup application has to be able interact with Nutanix system and command Nutanix system to take (and restore) snapshots. CommVault has the ability to use Nutanix native snapshots with Intellisnap capability. With CommVault you can take agentless backups and use Nutanix snapshots (and snapshot replication, since snapshot alone isn’t really a backup).
More info on CommVault Intellisnap for Nutanix:
Rubrik is currently supported only with Vmware, but I have understood that the next version (soon to be released) will support AHV as well. I have POC planned for Rubrik with Vmware and AHV, will report results later.
With other backup applications you will have to install backup agent for each VM running under AHV and then backup each VM separately as guest backups. No nutanix snapshots are taken during backup and cannot be used to recover data. For example Veeam has shared some information on how to backup AHV based virtual machines.
More info on Veeam / Nutanix AHV backup:
Most of the modern backup applications have support for vmware snapshots, so you have more options for backup. On the other hand many backup products for Vmware rely on changed-block-tracking functionality which has had integrity issues, like this. So while the functionality is better or you have more choices with backup applications, it isn’t a totally risk-free solution.
Are all snapshots created equal? Heck no. AHV is using Nutanix filesystem snapshots which are pointer based and do not have performance penalty, i.e no performance issues while writing, like copy-on-write-snapshot (most of the legacy storage arrays) or performance issues while removing snapshots, like Vmware snapshot which will freeze the original vmdk-file and collect changes into delta files, once snapshot are deleted changed blocks have to be copied from delta files to original vmdk-file. The longer you keep vmware snapshot, the more changes are in delta files and the longer it takes to remove the snapshot.
Nutanix has good snapshot and replication (both asynchronous and synchronous) capabilities, so it is feasible to solve issues related to data protection without any third-party backup software. Either by replicating data to other physical cluster or to cloud (AWS and Azure).
Third-party backup software is required for tape backups, but many organizations are protecting their data without tapes or are using multiple solutions for data protection. Typical example would be to use Nutanix native snapshots and replication for short-term backups and maybe run a monthly tape backup with separate backup application. This way “Fresh” data is contained within snapshots or replicas and can be recovered in less time than with tape. Stale or old data is rarely restored and in most cases restore of such data isn’t so time sensitive and can be accommodated with tape backups.
Nutanix supports Nvidia Grid GPU graphics cards on certain node models. GPU cards are typically used in VDI environments where the normal virtualized display adapter doesn’t have enough oomph. Examples of such environments are 3D or CAD design applications.
In order to utilize GPU cards in virtualized environment device drivers are required for the hypervisor in use. Currently Nvidia Grid cards are only supported with Vmware.
There are rumors about KVM (AHV is based on KVM) drivers for Nvidia Grid cards. Once Nvidia releases them, Nutanix can implement missing bits and provide GPU support also on AHV platform.
Affinity and Antiaffinity rules
Affinity rules are part of Vmware resource scheduler and with these rules one can make rules for virtual machine placement. Current version of AHV has mechanism only for initial placement, not active load-balancing based on resource consumption. Affinity rules can also be used to resolve some licensing issues, to limit running software only on certain nodes and thus having to pay software licensing only for the nodes used to run the software.
At the moment Vmware has better functionality, but the next release of Nutanix software will introduce new Dynamic Resource Scheduler (DRS), which has clear improvements over Vmware capabilities.
More info on AHV affinity rules:
Nutanix is one of the only Hyperconverged to solutions to offer synchronous replication over geographical distance. With sync replication/mirroring you can implement Metro availability and achieve zero recovery-point-objective (RPO), i.e no data-loss in case of site failure. With new witness or tiebreaker capability you can automate site-failover and reduce recover-time-objective (RTO), i.e time to start services on surviving site.
It is my understanding that Metro Availability functionality is currently supported only with Vmware. However with upcoming AHV version with improved DRS functionality resolves some of the hurdles related metroclusters and I wouldn’t be surprised if Metro Availability support for AHV would follow soon after.
More info on Metro Availability witness functionality:
Vmware has some functionality on changing the identity of VM and guest operating system automatically while cloning virtual machines. With AHV you have to use custom scripts to make changes while cloning machines, either with CloudInit for Linux or SysPrep with Windows.
Open vSwitch Management
Nutanix uses Open vSwitch under the covers to implement network functionality. With current AHV version management is mostly done via CLI, only very limited functionality in GUI. Next software version will add more network functionality into the GUI.
Thanks for reading, comments are welcome