VMware vSphere 6.5 – Is it all it’s cracked up to be?

I won’t keep you in suspense.  VMware vSphere 6.5 is good.  It’s really good.  Packed with new features and updates, many of which people have been asking for since forever.  It’s certainly not all roses and rainbows here but the list of negatives is far outweighed by the positives.  If you want the VMware marketing take you can check out the “What’s New in VMware vSphere 6.5” PDF they published recently.  Many sites have published articles on a lot of these features already.  I’ll try and not be redundant and give you my quick thoughts on the features I find the most interesting.  This may be a bit of a read, so I won’t waste space reposting images from the What’s New.

VMware vSphere 6.5 – The Cons

I figure let’s get the ugly out of the way before we start with the awesome.  This is a really short list as I mentioned before.

• Compatibility with other VMware products
  • There are a few VMware products not yet compatible with vSphere 6.5 as of the writing of this article.  VMware NSX is on the list and you can check out the others here.  I’m sure VMware is working on getting those products compatible so this is really a minor complaint.
• Compatibility with some hardware
  • There are still a ton of older and newer hardware & servers that aren’t yet compatible with vSphere 6.5.  This list is definitely going to change as support is verified but it’s certainly something you’ll want to check before upgrading.  As an aside I recently installed vSphere 6.5 onto a Dell R710, which is currently unsupported, and it works fine.  I did get a new message indicating the CPU may not be supported in a future version though.
• Log Insight 4.0 – Single Host
  • I was recently introduced to Log Insight and I’ll admit it’s interesting, with it’s integration with vROPS among a number of notable improvements.  It also now comes free with a vCenter license and grants you 25 OSIs.  The FAQ indicates multiple hosts can be assigned licenses but when I added a vCenter it said this license only works for a single host.  The message appeared intentional so I have to assume this changed from vSphere 6 and Log Insight 3.x or it’s always been this way.
• vSphere C# Client Removed
  • So it’s finally gone apparently.  I’m in the camp that believes this was a bad move.  I understand the need to consolidate development and to modernize the interface.  That being said the venerable C# Client, vSphere Client, Thick Client or whatever you want to call it, for most engineers and administrators is still way better than the vSphere Web Client or the HTML5 Client.  It’s faster and organized in a familiar way since it’s been in use across every version of vSphere until now.  The fact of the matter is I can build and configure an environment much faster with the C# Client than I can with the vSphere Web Client or the Host Client any day of the week.  There’s also an argument to be made that we should have some redundant capabilities for a management interface of vCenter and ESXi.  Even if it’s basic and not full of the advanced features, I’d still like it as an option.

VMware vSphere 6.5 – The Pros

Now for the good stuff.  I’m going to throw my 2 cents at most everything on the What’s New doc from above.  So I apologize ahead of time for being long winded here.  I will be skipping some things that I just don’t find that interesting.  So let’s get started.

• Migration Tool built into the VCSA Installer
  • Migrating from Windows vCenter to VCSA right from the Installer is a good change.  In a lot of instances the de facto upgrade method has been to rip and replace because it was easier than trying to upgrade.  The inclusion of Update Manager into the appliance now makes this a very attractive option.
• Appliance Management Improvements
  • You can now see CPU, Memory, Network and Database stats without having to use the CLI.  The interface for this is nice looking and really clean, I like it.  A nice addition to be sure.
• VCSA Native High Availability
  • Finally! It’s been a long time coming.  VMware has pretty much had nothing available for vCenter on Windows or the appliance, except encouraging backups.  They had that vCenter Heartbeat thing a while back but it didn’t catch on and it was put out to pasture in 2014.  The new VCSA HA feature uses 3 nodes, an active, passive and a witness.  It can be enabled, disabled or destroyed at will.  vCenter can implement all this on it’s own, no extra configuration or servers or other components needed.  This is by far one of the best new features.
• vCenter Native Backup & Restore
  • You can now backup the VCSA and PSC appliance from the VAMI or API.  I’d be surprised if someone didn’t already have something like Veeam backing these up, but it gives you another restore option if anything.
• vSphere Web Client
  • Each new version comes with performance enhancements to the vSphere Web Client and vSphere 6.5 is no exception.  There are also a number of organizational changes that should help ease the transition from the C# Client.  They’ve renamed several tabs and moved components so it’s more in line with what you’re used to seeing in the C# Client.  A nice thought but still not enough I’d say.
• vSphere Client (HTML5 Client)
  • The HTML5 Client is now built into the Windows and VCSA versions.  It’s fast and looks really slick.  It’s STILL not full featured and still not quite as good as the C# Client but it’s getting better.
• Content Library
  • I like Content Library quite a bit, unfortunately it seems to be one of the better underutilized features out right now.  Replication between Content Libraries is so easy, but most people still create ISO and Template datastores and manually copy contents to other clusters or remote sites.  Content Library may now see an increase in usage since it can be used to mount ISOs, update templates and apply Guest OS Specs.
• vSphere Update Manager
  • Update Manager is included and enabled by default on the VCSA now!  I can finally recommend the VCSA without any caveats.  Since it’s on the VCSA it’s also able to take advantage of the resiliency of vCenter HA.  No more additional Windows License and SQL database to worry about.
• AutoDeploy now has a GUI
  • Auto Deploy can now be managed through the vSphere Web Client.  I haven’t used Auto Deploy much but anything that makes things easier is certainly welcome.
• vRealize Operations Manager 6.4
  • vROPS 6.4 now has a an Overview Dashboard that contains some environment based information on components in the environment.  It also has some new views for each object and displays VM folders in Environment View.  I wish they would fix the way licensing is handled though.
• VMware PowerCLI
  • PowerCLI is now completely module based.  There’s also several updates to existing modules including the Core, Storage and Horizon modules.  For those that use PowerCLI often I can guess this is good news.
• VM Encryption
  • VM’s can now be fully encrypted!  The encryption happens at the hypervisor level completely outside of the VM.  This works with any guest OS as well.  One big caveat is you have to have a 3rd party Key Manager to make this work.  Cool idea, I’m having a hard time finding a good use case though other than checking a compliance box I guess.
• Encrypted vMotion
  • vMotion encryption can now be applied to VM’s in transit.  Sounds interesting enough.  Is this supported on Cross vCenter vMotion or Cross Site vMotion?  The PDF doesn’t appear to indicate it is.  Between hosts in the same cluster I don’t really see this as necessary since your vMotion network is likely on it’s own somewhat segregated VLAN.  I can see this as highly necessary on Cross vCenter vMotion or Cross Site vMotion though.  I guess we’ll find out the extent that this works when more information is made available.
• SecureBoot for VMs and ESXi
  • With all the boot loader craziness out there being able to use SecureBoot for VM guests and ESXi hosts is very welcome news.  For ESXi hosts there may be a few problems moving forward though as MANY manufacturers have unsigned VIBs that need to be installed to take advantage of on-host hardware or 3rd party MPIO which means you wouldn’t be able to use this at the host level until they get signed VIBs.
• Enhanced Logging
  • Audit level logging is finally here.  I know I’ve been asking for this forever.  Someone changes a VM vNIC to another network or changes vCenter or host settings. you can now see exactly what settings were changed and what the previous state was.  You’re going to need a Syslog server of course, maybe something like vRealize Log Insight.
• Proactive HA for Cisco, Dell, HP so far
  • If you have Dell Openmanage, HP Insight Manager or Cisco UCS Manager, they can now pass hardware based alerts to vCenter.  This data is then used by Proactive HA when a hardware component is degraded and it can put a host into Quarantine Mode where VM’s are migrated to healthy hosts, hopefully before hardware failure occurs.  I really like the possibilities of this feature, although I wish there was a VIB based option instead of the external 3rd Party management platform requirement.
• HA Orchestrated Restart
  • A similar feature has been a part of ESXi in a very limited fashion for a long time.  You could manage exactly the order VM’s spin up whenever a host is powered on.  This feature breaks when part of a vSphere cluster though.  With HA Orchestrated Restart we can now do this across a cluster through HA.  It’s about time.  Gone are the days of HA randomly spinning up VM’s due to power loss or outage scenario.
• HA Admission Control Improvements
  • HA Admission Control now defaults to a Cluster Resource Percentage policy which is the calculated percentage of total available CPU and Memory resources in the cluster.  This percentage is calculated automatically using the number of Host Failures To Tolerate (FTT) and assumes the host with the most resources could be the one to fail in a worst case scenario.  This is pretty interesting because many vSphere clusters contain hosts with varied amounts of resources.
  • The second enhancement is called Performance Degradation VMs Tolerate.  Not sure why they didn’t call it “Performance Degradation of VMs to Tolerate”, but no big deal I guess.  This percentage setting indicated what amount of performance degradation you’re willing to suffer in the event of a host failure.  Potentially useful, but we’ll have to see how well it works in the real world.
• HA Support for vGPU
  • This feature is a weird one to be sure.  vGPU enabled VM’s don’t support vMotion because of the tight integration with the local GPU.  With vSphere 6.5, vGPU enabled VM’s are supported by HA.  In the event of a host failure those vGPU enabled VM’s will now be restarted on hosts with available vGPU resources.  Sounds reasonable enough, but it’s really not.  In most vGPU environments you are taking FULL advantage of the NVIDIA GRID GPUs.  This essentially means you’re probably running the full amount of vGPU based VM’s on each host.  The density of VMs per host of course varies based on the vGPU profile you’re using.  The need to take advantage of the GPUs is largely based on the cost to implement NVIDIA GRID Cards.  Because of that cost it doesn’t make sense to have vGPU resources just sitting there available not associated to a VM already.  It’s sort of a puzzling addition, but I guess if you’re in a large enough vGPU deployment this may make sense.
• FT Improvements
  • Fault Tolerance now uses DRS to rank hosts based on available network bandwidth which should help for a more even distribution in heavy traffic environments.
  • They don’t explain how, but network latency between the primary and secondary VM’s has been decreased.  This should help adoption if it translates in real world usage.
  • You can now use multiple port groups to increase overall bandwidth available for FT logging traffic.  More bandwidth is always good.
• Predictive DRS
  • DRS can now leverage vRealize Operations Manager data to see when resource spikes occur and utilize that data to better balance the load across the cluster before those spikes occur.  The only caveat, if I would even call it that, is this requires you to have vROPS.  I like and recommend vROPS for most environments anyway so this is a good thing in my opinion and yet another welcome feature to improve DRS metrics for load balancing and distribution.
• DRS Additional Options
  • The VM Distribution option evenly distributes VMs across hosts in the cluster but still places priority on performance.  I guess it depends on utilization and what your comfort level is since DRS normally doesn’t kick in as much on underutilized clusters, but you still may want a cleaner balance between hosts.
  • The Memory Metric for load balancing option uses consumed memory instead of active memory.  DRS normally uses active memory + 25%.  Using consumed memory is of course going to have a higher metric but will account for memory utilization spikes in VMs and will sort of get ahead of the curve.
  • The CPU overcommitment option can enforce a vCPU to pCPU ratio at the cluster level.  This would certainly cut down on CPU Ready State issues but may be a bit of a drastic option and could really lower density per host.
• Network Aware DRS
  • DRS now uses network utilization as one of the metrics to determine if a VM needs to be migrated.  I’m surprised it’s taken this long for this to be added.  A great addition to DRS.
• vSphere Integrated Containers
  • Containers are here!  I really don’t have any experience with Containers although I’ve read quite a bit about them.  I think eventually they’re going to be something you can’t do without.  At least in my circles I don’t see anyone using them yet.  I’m hoping to build the components for this in my lab and really see what this is all about.  Perhaps more to come later.
• Advanced Format and 512e Disk Support
  • VMFS 6 available with vSphere 6.5 supports 512 emulation mode (512e).  This feature will provide support for Advanced Format drives and storage arrays that support 512e LUNs.  Both are going to become more prevalent as drive sizes increase.
• Automated UNMAP
  • UNMAP is back and now it’s Automated!  Originally introduced in vSphere 5.0, the UNMAP VAAI primitive would free space deleted from a VMFS datastore on the storage array LUN.  It was quickly disabled due to performance issues and reintroduced in vSphere 5.5 as a manual process you had to run from the CLI.  The current iteration requires VMFS 6 and vSphere 6.5 of course.  It also runs in a much lower performance context which should alleviate the previous performance issues.  Time will tell how effective it is.
• LUN Scalability
  • The upgraded support for 512 LUNs and 2000 paths is nice.  Not much to say here other than this will add a bit of scalability to your storage infrastructure as remarked in the PDF.
• iSCSI Static Routing
  • So we can add static routes between an iSCSI target and initiator.  I have a hard time seeing a use case for this since you normally want to keep the host initiator and target on the same subnet.
• Dedicated Gateways for VMKERNEL ports
  • Last but not least, VMKERNEL ports now support dedicated gateways.  Depending on your network infrastructure this could definitely be useful.

That’s all I got for now.  vSphere 6.5 is for sure worth checking out.  Loads of new features and capabilities.  There are a few issues of course but they’ll get ironed out over time I’m sure.  As far as production adoption goes I generally say wait till after the first large update is released.  Thanks for reading!