ra2 studio - Fotolia
Early in February, VMware CEO Pat Gelsinger talked up the impending release of vSphere 6 and its more than 650 new features, but also took the opportunity to refine the software-defined theme VMware has been touting for the last few years.
Gelsinger said VMware is the company that can help IT become more flexible by using such technologies as Virtual SAN to manage storage needs and NSX to make the network more malleable. Gelsinger also announced the release of VMware Integrated OpenStack, the company's first product on the OpenStack platform that is free to vCloud Suite and vSphere Enterprise Plus customers.
SearchVMware asked its advisory board members for their thoughts on vSphere 6 and the other recent developments from the virtualization company.
While many of the improvements in vSphere 6 are targeted at folks running massive data centers, there's quite a bit to be excited about if you have a smaller IT shop like I do.
My favorite new feature is probably an obscure one. The vCenter Server Appliance (vCSA) will have full parity with the Windows-based vCenter Server, able to support 10,000 powered-on VMs and 1,000 hosts, all with the included PostgreSQL database. This is great news; it can save a Windows license and simplify management by not requiring Oracle or Microsoft SQL server. Being a virtual appliance, it installs faster and easier. Our organization started using the vCSA about a year ago, and it's been great for us. We've gotten better performance with less configuration and less stress because we don't have to worry about Microsoft patches.
As an academic institution, we've made a large investment in virtual desktops across campus. Something that's proven difficult for us to do in a fiscally responsible way has been to support engineering applications like AutoCAD and SolidWorks over virtual desktops. The nVIDIA Grid vGPU support in vSphere 6 is a huge win for us, because it allows the VMs to access the nVIDIA GPUs using native drivers. We've had this beta code for a while, and our testing has been very promising. This feature would allow our students and faculty to use complex applications outside of the physical computer lab and do their work from anywhere at any time. This is the technology that I'm most excited about because it's directly visible to our users; they can see the improved technology right in front of them.
While we have just about all of our mission critical applications virtualized, we haven't used Fault Tolerance (FT) because, until now, it's only been available for one processor. Having Fault Tolerance available for VMs with up to four CPUs makes the powerful FT technology something we can actually use with systems that are critical to our operation.
Like most organizations, we've got a ton of security policies that we need to adhere to, and our security team doesn't always appreciate the difference between physical and virtual systems. VMware has made some useful security advancements in vSphere 6. Most notably is passing the vSphere username into the ESXi logs; this means we no longer need to correlate vSphere logs and ESXi logs to find out who performed a certain action. Now we can see the username of the vSphere user right in the ESXi logs. It's not sexy, but it's a great feature for audit compliance.
Now vMotion is capable of functioning across a 100ms round-trip latency link. It's a sweet feature. I can't use it because we don't have any overseas data centers, but it's definitely cool that you can have that level of application portability if your business can utilize it.
Speaking of things that I can't utilize, and since VMware loves to flaunt scalability, it's probably worth mentioning some improvements there. VMware doubled maximum cluster sizes from 32 to 64 and increased VM maximums to 128 vCPUs and 4TB of vRAM, and you can push up to 480 CPUs and 12 TB of RAM in a host. Of course, for those of us that aren't running multi-billion-dollar data centers, this is an academic increase. I wasn't constrained by vSphere 5.5's 320 CPUs per host. However, I do like knowing that my paltry 512 GB RAM hosts with 24 cores is way under a supported maximum. It's kind of how I feel when I'm on an elevator; I may not weigh 8,000 pounds, but I'm glad the elevator in my office building can hold that much.
All in all, I'm excited about vSphere 6. It may not be revolutionary, but the evolution of the centerpiece of VMware's offering is always cause for some excitement.
Personally, I'm really excited about VMware Integrated OpenStack (VIO). Making OpenStack that much easier to deploy and maintain gives infrastructure time to give the developers the OpenStack capabilities they want. There is a bigger learning curve with OpenStack and setting up a separate environment would be time-consuming for the first rollout. VIO will give companies the ability to deploy it out quickly into a production-grade environment with rolling updates. That's a big win in my opinion. Plus, there seems to be a good amount of APIs that the developers can put to good use. Making it included with Enterprise Plus was a smooth move on VMware's part. They are staying in the game as people are looking at open source opportunities. Companies not paying attention to OpenStack and other open source platforms should start to look further down the road. The trend really is moving to more self-service and automation.
Another great announcement was VVOLS. It has been a long time coming. I'm curious to see if it actual makes it in vSphere 6 because this has been promised for about two years or so. VVOLs introduces storage policy-based management and eliminates the need to manage at the LUN level. Being able to manage and ensure storage performance per VM will be a positive feature for me. Granular control just gives more power to the administrator.
A lot of other great announcements were made as far as cross vCenter vMotion, linked mode for the vCenter appliance and instant cloning of a running VM. I could go on and on but lastly, I pray -- and I'm sure many other sysadmins do -- that the Web client interface is truly better. I still struggle with its slowness in 5.5 and often revert back to the thick client for quick operations. People will be more apt to adopt it with better responsiveness and the flattened menu options.
I think one of the best changes with vSphere 6 is it will refresh some of the features that were somewhat rushed into vSphere 5.5, such as VSAN. The new release should bring about a more mature software-defined storage technology. Companies using software-defined storage could find large savings when compared against the established SAN and NAS vendors. Over the long term, I expect software-defined storage to become the business norm.
One vSphere 6 feature that sticks out is the revised fault tolerance (FT). This technology promised so much from the start -- easier system management, no expensive cluster software -- but it fell short since it only supported one vCPU and small amounts of RAM. The use of FT was limited since most modern software recommends two vCPUs. With vSphere 6 supporting four vCPU FT, this is more of a usable technology. For those machines that need almost 100% uptime, but for companies that don't have the experience in IT or the money to do clustering, it will be an appealing prospect. The one caveat is FT works with only one machine, and it will need maintenance but this can be planned for off hours to minimize downtime.
Overall, I would consider vSphere 6 a good incremental update that fixes some of the issues in earlier releases of the newer technology components.
With vSphere 6 comes many technical enhancements that are good news for customers to improve performance, availability and usability in the data center. But there are also many improvements that administrators will like very much and will make life easier and allow for new possibilities.
First, there are enhancements to vCenter and single sign on that allow for a better distributed installation that will improve performance and availability of the management platform. Another important enhancement is the possibility to perform a vMotion between two vCenter servers. In combination with the longer round-trip time of 100ms it is not only great to be able to migrate virtual machines to another site but it allows for an easy migration of virtual machines from one vCenter inventory into another one, including migrating virtual machines from a Windows vCenter server to a Linux vCenter appliance. While it is not a true migration path, it does make the process easier.
We should not forget the promise of the improvements in the vSphere Web client. Most administrators will be skeptical -- and justly so. However, we're hearing login times that are up to 13 times faster, right-click menus that are visible and usable four times faster, and other actions that are now at least 50% faster.
One other thing that will greatly improve the efficiency of how administrators work with vSphere is the multi-site content library. This allows the storage of templates, ISO files and scripts in a central location that can be used from several locations and kept synchronized to keep the entire library consistent.
The big features that jump out at me for vSphere 6 are VVOLs, NFS 4.1 and four-vCPU Fault Tolerance (4FT). Project Fargo is cute -- and great technology, don't get me wrong -– but it isn't going to solve my problems right now. VVOLs, NFS 4.1 and 4FT are.
VVOLs means people stuck using block device storage like iSCSI, Fibre Channel or -- horror of horrors -- FCoE aren't entirely doomed to a life of mediocrity and tedium. As I like to say, "Businesses don't make money resizing LUNs." With VVOLs, companies that use LUN-based storage are one step closer to the ease of use those of us who use NFS have known for years. Assuming, of course, the array vendor supports it.
NFS 4.1 is the exact opposite. The biggest issue with NFS was VMware didn't support multipathing. With the LUN-based protocols, when a switch went down, the server could reroute through alternate links and your VMs would (mostly) stay up. This depends on a lot of factors, I've seen VMs not survive the iSCSI timeout for a failed link.
NFS is far, far easier to use than any of the LUN-based protocols, and you don't have to waste the time of your business resizing those LUNs. So wouldn't it be nice if you could have the best of both worlds? With NFS 4.1 you can: Great ease of use, no LUNs to fiddle with and multipathing redundancy. Assuming, of course, the array vendor supports it.
4FT is something I consider a transformative technology. In the old world of HA, when a server went down so did all its workloads. HA allowed those workloads to be restarted on another server immediately, but that's simply not good enough for the most critical workloads.
Old school fault tolerance only allowed for a single vCPU in VMs (1FT). 1FT VMs weren't really all that useful; there isn't a whole heck of a lot you can do with just the one vCPU. Most workloads that can get away with just one vCPU usually are just fine with HA. They don't need FT.
4FT changes all of this. 4FT means that real workloads can be set up to be fully fault tolerant. Walk up to a server and physically shoot it and the workload will carry on doing its thing on the other system. Assuming, of course, you have enough network bandwidth to synchronize four vCPUs between servers.
As with everything, there are caveats. That said, I feel that vSphere 6 is a major release simply because of the critical functionality it is bringing to meet real world needs. These aren't a bunch of features that need to mature a bunch before they're useful, or which are great ideas with horrible UI implementation. These are features that real sysadmins in real data centers will need and want today.
I like several of the new features in vSphere 6, but the reality is I don't see most customers being able to benefit from them for some time. My favorite items would be the vMotion enhancements, namely the capabilities for long-distance migrations, cross-vSwitch migrations and cross-vCenter migrations.
These updates will be attractive to customers who want additional flexibility to move workloads around in their environment. But they have to be on vSphere 6 to take advantage of them. There will likely be a six- to 12-month period for a substantial amount of customers to take on this type of upgrade -- and even longer for many large customers.
As far as the updates to maximums around VM size and other limit changes, that is less exciting to most customers and feels more like bragging rights in the marketing war with VMware's competitors.
Check out the VMware vSphere 6 storage capabilities everyone’s so excited about