Avoiding downtime with VMware Fault Tolerance and High Availability
A comprehensive collection of articles, videos and more, hand-picked by our editors
Now that the thumping techno music has faded and the vendor parties have wound down, it's time to evaluate the announcements from the VMworld in this VMworld 2014 recap.
VMware continued to trumpet its vision for the software-defined data center (SDDC). Perhaps the biggest splash in this area was the EVO announcement, which ties VMware software to partner hardware for a hyper-converged appliance. EVO:RAIL was positioned as "SDDC in a box," while EVO:RACK is meant for larger enterprise needs.
In cloud computing, VMware announced vCloud Suite 5.8, emphasizing better interoperability between the suite applications and the NSX network virtualization product. In a nod to the developer crowd, VMware also announced the beta of its own "hardened" OpenStack distribution called VMware Integrated OpenStack, as well as its partnership with Docker to expand the use of lightweight VMs on the vSphere platform.
In a repeat of VMworld 2013, vSphere didn't get too much attention, but there were some intriguing teasers for the next release -- VVOLs, enhanced vMotion, multi-vCPU Fault Tolerance, an improved Web client -- with an expected release in 2015.
SearchVMware asked advisory board members for their opinion on the announcements and updates at VMworld 2014.
I think EVO was one of the more interesting things at VMworld this year. It looks like VMware has put together a nice solution for the first release. They might not be able to compete fully with some of the other vendors yet, but VMware has put them and customers on notice. If they stay committed to EVO, I think we will see a big improvement by VMworld 2015.
The next best thing is an EUC-related product suite that includes AirWatch. VMware is hard at work integrating the EUC products for a more complete product suite.
It was great to see a lot of interest from customers around NSX. This should help VMware retain market share of the vSphere platform by building quality products on top of the hypervisor. It may take a while yet to see a lot of deployments, but I continue to see more customer use cases that could be a good fit for NSX.
For me, the big news was EVO:RAIL. Personally, I don't think this was a huge surprise. Anyone who has been to a VMUG meeting lately or had an opportunity to visit with vendors can see that hardware vendors were already well along this path. What makes this different is that it is coming from the software side, then to hardware, instead of the other way around.
An advantage that VMware has in doing this is they can make sure the appliances are supporting the latest versions and patches of its products close to time of release. As it is now, I have noticed that some convergent technology hardware vendors do not support the latest versions or patches in a timely manner, which is a security concern.
If VMware can keep the hardware vendors in-step with current releases -- adequate quality assurance from both is essential -- then I think VMware could have a winner here. The concern of vendor lock-in is a non-issue to me; it is more important that my hardware and software vendors not point fingers at each other. EVO:RAIL could be the solution to that.
Again, we have VMworld without a new major version of the vSphere platform being announced. Instead, we get the promise of a major version upgrade that is coming some indeterminate time in the future.
The big new thing for me was EVO:RAIL with VMware enabling its hardware partners to enter the hyper-converged infrastructure market.
Apart from renaming products, the other major news is better integration between the on-premises vSphere and VMware's own public cloud, now renamed vCloud Air. The improved integration between on-premises vSphere and vCloud Air will help customers migrate workloads to VMware's own cloud. Will the same integration help customers move to a public cloud that VMware [does] not own?
EVO:RAIL is an interesting play that suggests that tightly defined hardware is an important part of a good hyper-converged platform. VMware has experienced some pain from having a loose hardware compatibility list for VSAN. There have been reports of one customer undergoing a nightmare recovery from a VSAN node failure, leading VMware to change its supported hardware for VSAN.
EVO:RAIL also makes deployment as simple as possible, removing initial setup as a barrier to entry. While vSphere with VSAN is not terribly complex for a specialist to set up, it is beyond many IT generalists. EVO:RAIL, on the other hand, could be deployed by someone without significant training. This simplicity of deployment is a trend that promises to make infrastructure as easy to use as public cloud infrastructure.
One of the interesting launches at VMworld was DataGravity, who launched its own storage array. The unique position for DataGravity is that the array will analyze the content of the data being stored, cracking open file systems and even databases stored inside VMs. Once the data has been analyzed, it is searchable, even by end users. This could transform the way we handle semi-structured data, such as Word documents. If the content of the document is indexed, then it is less important where the document is saved, since we can find the document by searching for the client or project it is about. Indexing data isn't new; any decent document management solution will index and enable searching. Placing the indexing inside the storage array is new and seems like a good idea. Arrays have ample CPU power and are very close to the data, reducing the storage network traffic for indexing.
[The fact] that vSphere didn't get much attention is probably a good thing, from my perspective. Products only get great publicity at big events like this if they're new -- and if they're new, it means they changed something. I'm change- and risk-averse when it comes to the platform 95% of my servers and applications run on. I don't want it to change that often. VSphere 5.5 is incredibly stable, and I like having only some incremental improvements. Having a three-year cycle for the software isn't a bad thing. Of course, all of that said, the teasers for 6.0 were great. There's a lot of very useful stuff coming, especially the longer-distance vMotion. I love the idea of having [the] ability to migrate running VMs across a long area. This is like 'insanely great' news for admins looking for some geographic resiliency for applications that can't go down.
EVO:RAIL is big news -- not because VMware is designing any hardware specifications, but because it confirms that the SDDC isn't about trying to make a bunch of disparate point solutions work together. It's about having the tools to not need to worry about the nitty-gritty technical how-to anymore and, [allowing] IT to spend more time serving the needs of the business. As a manager, the less time I have my staff futzing with all the components that make up the data center, the better. I can have them working with our customers to be innovative and work on projects that are strategic or differentiators in the marketplace.
VSAN is a similar story. The real beauty of it isn't the capital savings it provides -- though that's never a bad thing -- it's that I don't have to dedicate employees to managing a monolithic storage device. It is becoming less about what individual parts you know really well, and instead about what business value you can bring to the table.
VCloud Air is a compelling solution for companies looking to go to the cloud. For those customers, you get the option of moving your workloads that are already running on vSphere to the cloud without needing to redesign or rework anything. This is a big difference from other vendors in the cloud space. There's little more disruptive than needing to migrate an application to a new platform. With vCloud Air, I don't need to do that. I can move my VM, appliance or vApp up to the cloud when I need some extra capacity.
I didn't dive too deep into this announcement, so I'm not sure what the back end of all of this is, but I would assume it's running the same software that they sell to customers. That means VMware is running its own software at massive scale. As a customer, I like that. For instance, I'm pretty sure that a lot of Microsoft's cloud is custom-built instead of using its own software. Regardless, vCloud Air might be the best way for a company to start moving to the cloud, because the barriers are so low.
I kind of wish they picked a different name for vRealize. I think it's a pretty good suite of software hidden behind a silly name. I think vClarity or something would have been better.
What really interested me was the push in end-user computing to tackle the workloads that have been stubborn holdouts for a while. One of the use cases at my day job is trying to virtualize some teaching labs in the college of engineering with software, such as SolidWorks. This is a tough application to virtualize a lot of -- density is an issue and the graphics are killer. There have been various attempts to help, such as vDGA and vSGA, but the hardware has lagged, and it's not a perfect technology to begin with. Being able to cost-effectively move these kinds of apps with performance parity to the workstations the users are used to is massive. For many organizations, engineers are where the patents are, before they're patented. Keeping that kind of data secure in the data center should be a top priority, and this is an easy way to get there. Of course, running SolidWorks on your iPad is pretty cool, too.
VMworld to me is not about the announcements or new features. It's really all about networking and getting to see people face-to-face that I've spoken to over the year through social media outlets. So did it meet my expectations? Absolutely.
That said, if I had to rate VMworld by the keynotes, announcements and releases, I would definitely say that this year was a little lacking. This is the second VMworld that we've been through in which we haven't seen a major vSphere release. The yearly release cycle of its core hypervisor offering was something that I have always looked forward to and [that] always generated a lot of buzz. Perhaps the biggest miss in terms of announcements though was the absence of a general availability of Vvols [VMware Virtual Volumes]. Vvols is something that VMware has mentioned for the last three VMworlds, and we are still waiting to see it go GA.
Now that's not to say VMware didn't have a lot of new items to talk about. EVO:RAIL seemed to steal the show in terms of creating conversation amongst vendors and attendees. EVO:RAIL certainly is poised to change the way we architect data centers.
Again, my expectations for VMworld are not about announcements. It's all about community. And VMware never fails to provide with opportunities for networking. The Hang Space setup was great. The #vBrownbag Tech Talks were awesome. And the breakout sessions were amazing in terms of getting deep-dive knowledge on different solutions and products.
In vSphere, multi-vCPU fault tolerance is an absolute game changer -- if VMware can pull it off. It's technically difficult and unbelievably resource-intensive, but it would also be worth the vSphere license on its very own.
The Web Client was sent from hell to make us miserable, and nothing I've seen so far indicates VMware is ready to listen to its user base and solve the very real problems with it -- or to simply carry forward the fat client for those of us Luddites who prefer to rock it old school.
Enhanced vMotion is something I'm less sure about. I need more details before I get excited, and I need to see it in action on my various workloads.
Vvols seem cool if you're still using LUNs, but all of this is still 'beta, beta, beta!' VMware had a VMworld -- and vSphere didn't show up. While I admire that they're willing to take the flak for that and ride it out until it's polished, by the same token, I'm not ready to jump on the hype wagon for something that isn't in my hands yet.
When you look at the rumors of MARVIN I had been hearing, and then compare it to the delivered product, all I can say is, 'What an overwhelming disappointment.' VMware cloned Nutanix, complete with tying VSAN to a pre-canned hardware config. Congratulations. There is no magic additive to make it work better, and VMware has managed to poison its relationship with its own ecosystem partners who are now terrified of getting Nutanix-ed.
Other companies are ready to deliver on the promise of true software-defined infrastructure, and it is now looking like they'll get there before VMware. That could be disastrous, or it could trigger a flurry of hyperconverged competition that drives innovation forward and prices down. VMware failed to deliver a killer product while alienating some of its most valuable allies. Bearing the past couple of years in mind, I'm leaning towards EVO as more "disastrous" than "game-changing."
Don't overlook the importance of one key announcement at VMworld: VMware's vCloud Connector now does layer 2 extensibility. This is huge, and it makes VMware's hybrid cloud offering the easiest to use in the industry. VMware has the right product, but -- as is usually the case -- at the wrong price. What is preventing uptake is two things: widespread availability of partner data centers outside the U.S. run by non-US companies -- zero U.S. legal attack surface -- and affordability.
VMware's price to service providers is too high for them to offer a VCHS-like offering for anything other than a premium price, and VMware itself wants a premium price for its offering. Data sovereignty and cost issues will keep this an enterprise-only affair for some time to come. A shame, since it's the best of the best and far ahead of the rest.
AirWatch is great, but do remember that it doesn't have much in the way of management for Windows or OSX desktops. For example, it doesn't do patch management. VMware has products that fill all niches for endpoint management and delivery. But when you add this SKU to that SKU and throw in this other SKU over here, soon you're so far beyond the cost of buying your workforce a really nice notebook -- or two, or three -- every single year, that it's time to start re-evaluating why you're doing VDI at all.
VMware has great technologies here. They are a purchase of Liquidware Labs away from having the best technologies here. But [it] needs to rationalize and consolidate its offerings into more easily consumable licensing -- and something that those beyond the Fortune 2000 can afford.
What caught my attention was that VMworld 2014 was functionally a VMworld where VMware forgot about the VMworld. The announcements were thin on the ground -- both from VMware and from partners. VMworld itself could have been better described as 'storage world.' There wasn't a whole lot of 'virtualization' anything there, just storage, storage, storage.
If VMware wanted to focus its energies to make VMworld itself better, it would be well-advised to both expand the role of the community and to raise the focus on the partner ecosystem. We live in a world where hands-on-labs can be done over the Internet and training for any technology you want can be consumed in the same fashion.
What is hard to do is get the top nerds from multiple companies together in one room to build a multi-company technology demo, show how its bits work -- and why -- and take questions from the floor. Flying around the world and staying in some of the most overpriced hotels on the planet for a week just to consume a few talks about new features is madness. Flying around the world and staying in overpriced hotels to see with your own eyes technologies from multiple companies working together to do things you didn't think was possible before -- that's worth it.
The real benefit to a conference is the bringing together of the greatest minds of an industry. I realize VMworld is a VMware affair, but if all it's going to be about is VMware itself, why do we need a conference? They can just form a queue outside the recording booth in Palo Alto and have an endless line of product managers record an endless stream of videos to flood YouTube.
People and products, not propaganda and promises. VMware is where it is, not because VMware itself is great, but because of the ecosystem and community that accreted around it. I feel like that has been forgotten this year, and the result was a VMworld far more limp than the year before.