vSphere 5 is dead, long live vSphere 6 – Virtualb.eu’s overview of the new platform.

VMware-vSphere-6-300x218VMware today announced  the new version of its flagship product: VMware vSphere 6.  According to my sources the GA date off the product will be March 2015.

It has been since July 12, 2011  that a major release (vSphere 5 – not counting vSphere 5.5 as a major release) of VMware’s flagship product has been made available to the public. In this article I will go over some new features and updates that I think are major game changers for the platform along with a general “What’s new” list.

Installation and Upgrade path from vSphere 5

• VMware seems to have listened to customers and worked on the capability to automate and customize the vCenter installation procedures to allow for customization during and after installation (the ability to use more custom TCP/IP ports, custom paths, uninstall and error messaging improvements. There is now also a guided installer for the vCenter Server Appliance (vCSA) which allows you to have some customization of the installation process. Just in case you are worrying that the installation process now will become a lot more difficult: don’t worry, VMware provides both ISO’s for guided (or simple) installation and scripted installation.

Virtualb’s take

I have worked in large environments on several occasions where scripted deployments of vCenter and all it’s components were mandatory for a multitude of reasons including avoiding vCenter servers with different configurations around different geographic areas, taking away the complexity of a customized installation in order for the L1 teams to perform the installation without escalations, uniformity of the global environment and compliance to global security policies. In vSphere 5.5 there was little customization possible and the installation guides for scripted installations were not always correct, a lot of custom hacking needed to be done to achieve a completely scripted build of vSphere. If the above holds true I welcome it as a serious enhancement and hope it will make our lives easier scripting and deploying vSphere instances.

Extended support

• As usual in a new release VMware comes with expanded support for new hardware, and guest operating systems:

  • Additional support for Free BSD 10.0 and Asianux 4SP3 as guests
  • VMs will scale up to a monstrous capacity of 128 vCPUs and 4TB vRAM
  • Support for a cluster size of 64 nodes and 8000 VM’s* with monster hosts running  a maximum of 480 physical CPUs, 12TB RAM and 64TB of storage
  • Distributed vSwitch improvements enabling service level management for storage I/O, allowing network I/O control for each VM, and also for the network I/O of the distributed vSwitch bandwidth.
  • Directly linking to the above feature VMware now supports VMs with up to 4 vCPUs with vSphere FT (Fault Tolerance). I am assuming here that network I/O control will be available to control the bandwidth on the replication traffic for the supported VM’s as we now FT has a very high bandwidth requirement even with one vCPU as CPU instructions and Disk IO needs to be mirrored real-time to the secondary or standby VM.
  • vMotion is now possible with a network latency of100ms per round trip and the boundary of the vCenter as VMotion encapsulation has been lifted enabling workload shifting between virtual datacenters as long as they are on the same L3 network of course. There is now also support for vMotion across virtual switches which allows easier management functions and smoother migration paths.

Virtualb’s take

It’s the birth of the monster hypervisor! The with these hypervisor capabilities we can say the rise of the AFA’s (All Flash Array is now backed by the CPU and memory that can take full advantage of these super fast and resilient arrays. Old spindle arrays won’t be able to deal with the IO’s that can be served from these mean machines and I doubt hybrid arrays will do.

Finally support for 4vCPU FT VM’s opening up this feature for use in the catalogue offering (1vCPU vm’s I haven’t seen in a while) Considering the massive band with requirements I talked about before I believe this will be a feature that could come in handy within data centers over virtual clusters or even virtual datacenters as the vMotion restriction between virtual datacenters has been lifted. (Unless the CPU instructions and disk IO has been severely compressed for WAN optimization.)

Yay for the vMotion enhancements: This will be hit it home with L1 and L2 support teams globally, making their jobs easier moving VM’s around as updates or upgrade are performed on server HW or software. Lifting the vMotion round trip on the other hand will allow for easier cluster management and better disaster avoidance (DA) configurations – assuming you have storage replication between sites.

Storage

• Virtual Volumes (VVol’s) – Long promised (like multi CPU FT) and finally arrived: Virtual Volumes. In essence this means that existing SAN/NAS array will become VM-aware making the    VMDK a unique object on the storage array rather than an object in a datastore that is defined by a LUN on the array. To support this new feature a new set of API’s for data operations at the  VMDK level have been made available. Note that your HW vendor will have to support this feature and will require a (non)-disruptive firmware update to enable this feature, depending on your HW vendor. More information on the availability of VVOLs for your array on The Register (http://www.theregister.co.uk/2014/07/22/gearing_up_for_vvols_vmware/)

• Virtual Storage Area Network (Virtual SAN) v2.0 will be released as part of vSphere 6.

• NFSv4.1 with Kerberos authentication.

• Additional support for Storage IO control allowing per Virtual Machine storage reservations to guarantee minimum or maximum service levels.

Virtualb’s take

Finally VVOLs become a reality: this will enable VMware administrators to greatly simplify in the deployment, automation and monitoring VMs as vCenter will directly be talking to the storage array and VVOLs will not require the management of LUNs on the backend array as each object will be created individually on the backend array. Great news for management and monitoring of VMs on the vCenter side as this will give the VMware administrator a direct look into the array greatly simplifying monitoring storage performance.

VSAN 2.0 – As VSAN 1.0 was only meant for DEV/TEST workloads (source VMware) VSAN 2.0 should be the complete release. If you have been following VMware releasing new features in the vSphere suite this is not a new one. Let’s see what the new version will bring or what I would expect: enhanced monitoring (not a separate DEV Tomcat based web tool but actual integration in vCenter), enterprise features like compression and deduplication, cheaper licenses rather than  single CPU hypervisors to cut costs and VM location awareness. I realize that VM location awareness is topic that is debatable but whatever doesn’t need to go over the network shouldn’t go over the network, that’s my opinion. Also an updated HCL with a wider adaptation of RAID controllers would be awesome but I don’t think that will be the case. Looking forward to see how far version 2.0 will scale when it comes to storage clusters.

ESXi NFS 4.1: This will technically enable active/active multipathing for NFS arrays (Parallel NFS)  from the ESXi server. This will require an NFS 4.x aware array. Possibly with new multipathing policies for NFS. NFS 4.x  requires a server to store metadata. Not sure which storage vendors support NFS 4.1 at this point but definitely a big step forward as this will greatly enhance the performance of NFS filesystems. More on PNFS http://www.pnfs.com/
*pNFS is currently not supported on vSphere 6 when using the NFS 4.1 client. (together with other features like SDRS, SIOC, SRM, VVOLs and NAS VAAI). In order to achieve more throughput session trunking can be used – see Chris Wahl’s article .)

Other features

• Multi-Site Content Library will enable you to define a shared library between vCenters that can contain the following objects:

  • VM Templates
  • vApps
  • ISOs
  • Scripts
  • Virtual appliances (OVF)

• Performance improvements to the vSphere Web Client including login, home page loading, action menus, related objects and summary views. The C# vSphere client is still available in this release and according to VMware this is the last time the client will be released. The C# vSphere client can manage hosts running hardware versions 8 to 11.

• Enhancements for managing VMware and third party certificates by providing a new command line interface (CLI) for managing both VMware and third-party certificates.

Virtualb’s take

The shared content library is a great idea, I would have called it a Distributed Content Library as I assume there will be a managed content library that will be replicated to other vCenters servers datastores. Again this will cost network bandwidth and disk space but in essence a good idea to keep configurations and versions of templates and ISO’s in sync between different geographic regions and different support teams.

The web client. I could tell a long story about this but I am not going to do it. Very simple: keep the C# client or change the webclient to HTML5. Do not keep using Flash for the web client as it is simply impossible for a lot of customers in high security environments to even run Flash. I would have thought this would have been picked up by now but the message seems to very slowly drip in hence supporting vHW version 8 to 11 again in the C# client 5.5u2b, a feature that was removed from C# client 5.5. If you have managed 1000’s of VM’s and hundreds of hosts you know what I mean: it is impossible to do this job with the webclient.

Virtualb’s final take

A lot of new and long anticipated features and updates in this release of vSphere. A release who took a long time to materialize so hopefully this will equal a good release with stable code. It looks like VMware has listened to their user base and made some long awaited adjustments to the vSphere platform: I expect a lot from VVOLs, VSAN 2, NFS4.1 and the support for scripted installations. It would be great if VMware would come down from their throne and just keep the C# client around and recode the webclient in HTML5. Let’s hope the new features will integrate well with the other products in the VMware suite and I think we have the potential of a great release as most features have been on the wishlist of many VMware administrators for a long time.

Virtualb will be at VMware’s Partner Exchange 2015 in San Francisco next week and I hope I can bring you some more details on the above new features in follow up blog posts the coming week(s).

Disclaimer: In this article I have not mentioned everything that is new in this release, just the things I think jumped out. I refer to other documentation or blogs for a more comprehensive list.

 

*Updated after initial release

Comments

  1. You released this too early. Announcement does not take place for another 8 hours.

  2. WOW.. Waiting for the official release !!!

  3. Must be a typo. I think the config max for VMs per cluster is 8k. See ya

  4. Andres Rojas says:

    Thx for your post, you have summarized all the important/key info

Speak Your Mind

*

%d bloggers like this: