Rumors of a VMware hyper-converged infrastructure project first surfaced in March, when reports of something codenamed Project Mystic made the rounds.
While speculation was rampant, there was little evidence to confirm the existence of this construct. Then Fletcher Cocquyt , a technical architect from Stanford University, reignited the chatter after he tweeted a picture on June 6 purportedly showing a poster on the VMware campus with the text, "Introducing the world's first 100% VMware powered hyper-converged infrastructure appliance." The poster also featured the name "MARVIN" and declared, "Arriving summer 2014."
Some sleuthing uncovered VMware had filed for the MARVIN trademark on Jan. 8, 2014. In documents, the company described that the trademark was for, "Computer hardware for virtualization; computer hardware enabling users to manage virtual computing resources that include networking and data storage."
As of this article's publication, VMware has made no official announcement. Whether MARVIN is real or an elaborate fabrication, is this offering something the market needs or has another vendor already filled the niche? Is it another sign some administrators would prefer one throat to choke when problems erupt?
SearchVMware asked its contributors and other noted experts about the potential of a VMware hyper-converged offering and whether it makes sense for the virtualization company to enter the hardware business.
Does VMware have a hardware initiative?
VMware is a software vendor and will be for many years to come. Adding a hardware branch to the company is something not to be taken lightly, so the rumors about the trademark that VMware has registered for selling virtualization hardware must mean something else.
I think these are the two most likely scenarios: Either VMware has worked with an existing hardware manufacturer to develop an integrated server with networking and storage that can be used as a hardware building block in the foundation of the software defined data center (SDDC). Or it is an initiative where VMware will allow hardware vendors to sell such building blocks under an OEM license from VMware that will provide the integrated hypervisor, networking and storage software.
And this would make perfect sense, because software is becoming more and more important compared to the dominant role that hardware -- especially networking and storage -- has played in the past. So having easy-to-deploy building blocks would be a great solution to create the SDDC. And this is useful for "regular" data centers and data centers that are used to provide a cloud solution. Being able to easily and rapidly scale up the compute resources while staying within the same pool of resources -- because they are controlled by one software vendor -- is a great benefit compared to having to work with a mix of hardware and software vendors.
Read more form Rob Bastiaansen.
I think that when VMware enters the hardware space with a hyperconverged-type offering, it will prove that customers are serious about this type of deployment. VMware will likely be behind in features for a while since other vendors have been selling this type of solution for several years.
The most exciting part will be what VMware attempts to offer in the way of automation and services with this type of platform. By removing the complexities of historical enterprise infrastructure, they are opening up the possibilities for better automation and scalability. I don't expect everyone to run out and buy whatever MARVIN turns out to be, but there will be a tremendous amount of demand from customers to learn more about it.
Read more from Brian Suhr.
Sander van Vugt
The possible plans of VMware getting into the domain of hardware/hypervisor integration are understandable, but don't bring anything good to the market. It looks like VMware is trying to offer a complete stack solution. Oracle purchased Sun for that reason. As soon as a large majority of customers has adopted such a solution, there will be no limit on the pricing of these products anymore. Customers depend on it, so hey, let's just raise prices without offering any product solution in return. For that reason, I don't like it.
Read more from Sander van Vugt.
I have no knowledge of MARVIN or Project Mystic, [and] because I have no knowledge of the potential product, I can't really comment on what VMware is attempting to do. I can say that I think there is some inherent value to hav[ing] a single vendor provide your computing stack. More and more often, I think of hardware as being from the hypervisor down, and there's some value in being able to call a single vendor with any problem and know it won't get blamed on a third-party vendor. Every IT administrator has a war story where they were caught in a spitting contest between two different vendors, neither of which would accept responsibility for the issue and therefore withhold support.
While I don't have much to say with respect to MARVIN, I do have some thoughts on converged infrastructure in general.
As organizations continue down the virtualization and cloud path, and increasingly abstract the applications and data from the hardware, it only makes sense they would attempt to derive savings not only from higher consolidation rates, but also from a reduction in hardware maintenance and support costs. It's no secret that there's a whole category of IT professional that have made a career out of maintaining highly complex, vendor-specific IT infrastructures. These skills take time and talent to build and must be developed -- often at the expense of other skills or talents that could [make a difference] for an IT operation or a business.
Instead of investing all of this human and financial capital in hardware platform management, it makes sense to pick simple, scalable solutions that are designed for the solution the business has chosen. Because applications and data are abstracted from the hardware, there's no reason to have these monolithic hardware systems that [feature] all purpose-built hardware. Instead, because of the SDDC, there's no reason for vendor-specific platform management; you can run commodity hardware. If anything, what VMware has done very well is make the argument between which server manufacturer to choose completely irrelevant.
With the hyperconverged infrastructures like Nutanix or SimpliVity, you can take it one step further and not have to think about each area of your infrastructure such as storage, compute and networking separately. Purists will argue that you end up overbuying in one category or another because growth in all of these areas isn't directly proportional. I agree, but the extra spend in hardware is likely insignificant in comparison to what is saved in terms of design, engineering and maintenance. As an example, if it were more efficient to use separate storage, networking and compute nodes, then Google, which runs some of the largest technology plants in the world, would almost certainly use that model. Instead, at that scale, it makes the decision that much more clear -- the data center is changing irrevocably -- and simplicity is winning.
Check out Steve Athanas' Twitter.
MARVIN has certainly sparked a lot of interest amongst the VMware ecosystem. At this point, we have to take the rumors for what they are: rumors.
That said, when I think about what MARVIN could be, two things come to mind. Either it could be a piece of custom-built hardware containing all the pieces of compute and storage, bundled with vSphere and VMware VSAN to provide a hyperconverged infrastructure similar to that of Nutanix and SimpliVity. Or it may be a sort of reference architecture/hardware compatibility list that VMware would provide to its OEMs and partners, allowing companies like Dell, HP, etc., to provide a sort of build-your-own-hyperconverged infrastructure to customers.
In my opinion, the latter seems more likely given that VMware has long touted the status of being a software company and has been recently pushing the commodity of the vendor-agnostic SDDC. Whatever MARVIN pans out to be, whether it be hardware or architecture, it's clear Nutanix and SimpliVity are filling a gap in the market.
MARVIN could be a great solution for a remote/branch office to provide the protection and performance in a very small form factor. Also, will MARVIN be a component in the proverbial "year of VDI"? If MARVIN could provide the performance and capacity needed to run a definite number of virtual desktops, with the simplicity of scaling out nodes to increase that number, I could see adoption in the VDI market growing -- providing the price is right.
The sizes of the Nutanix/SimpliVity booths at VMworld are proving that the convenience and simplicity of a hyperconverged scale-out infrastructure is gaining more traction in the enterprise, so I'm not surprised to see the company that provides the hypervisor with vSphere and the virtualized storage with VSAN, attempt to move into that market. It sounds like we will just have to wait -- probably until VMworld in August -- before we get any details.
Read more from Mike Preston.
It seems ironic to me that the biggest proponent of virtualization to reduce hardware needs could feasibly start selling hardware. Upon further thought, it makes a lot of sense.
I recently attended the VMUG conference in Denver, and it is clear that convergent virtualization technologies are overtaking the more traditional infrastructure. The simplicity and apparent ease of installation and maintenance make a compelling case for convergent appliances.
Is the field as well developed as it could be? Hardly! I believe that there is still room for new players in the convergent technologies arena. For example, at the conference were a number of convergent technology vendors. Yet, the price point seemed pretty high for what was offered, and some of their products did not support the latest software from VMware, which was a bit disturbing to me. I believe that if VMware was to enter the hardware market with convergent appliances, they could provide some options that competitors cannot, at least not in a timely manner.
Please consider, for example, what is required when a new software update becomes available from VMware. Best practice is to verify its compatibility with your hardware, which in itself can be daunting considering the number of components in a traditional infrastructure. As I witnessed firsthand, even convergent hardware solutions require additional time to certify the software on their platform. Now, if VMware were to offer a convergent solution, it could feasibly be fully tested for compatibility at all levels, and would be immediately available on their own hardware at the time of software release. Furthermore, if something did go wrong with it, you would not have the issue of different vendors pointing fingers at each other, which is very frustrating for the customer who is facing downtime and needs a solution fast.
If VMware were to make a convergent solution at a reasonable price point, I see no reason why it would not succeed.
Read more from Mak King.
I am leery about these MARVIN rumors. On one hand, having a reference platform that VMware can sell makes perfect sense. They have a sophisticated and capable sales force that does a lot of consulting with the kinds of enterprises in which bitter rivals, Nutanix and SimpliVity, are gaining wins. On the other hand, why would VMware -- a software company -- get involved in the messy details of hardware? It's not impossible. Other software companies do it, but they range from Microsoft, which does hardware badly, to Oracle, which does it slightly less badly. There aren't many stellar examples of the merging of software and hardware cultures at the sales end.
More likely, I think that MARVIN is a joint operation. Hardware supplied by HP or Supermicro, sales supplied by EMC. VMware has the right people and the technical knowledge to do the validation testing and deep out-of-the-box integration. In a pre-canned package, you'd want all the hardware sensors reporting, you'd want Intelligent Platform Management Interface (IPMI) integrated. There's a lot of bits to get just right. All companies involved throw some marketing dollars at it and presto! A new product is born.
This sort of joint operation would allow VMware to do what they do best: Make software work. EMC can handle the logistics of getting boxes, getting the software on them, testing and shipping them at industrial scale. Why would VMware reinvent the wheel there?
Do I believe MARVIN is a hyper-converged play aimed at Nutanix? Yes [and] probably Cisco as well. NSX is ready to rumba. With Cisco's acquisition of Whiptail being a poke in the eye for EMC, it's time for the VCE partnership to fade away. If MARVIN doesn't exist, then the outright hostility from leading community members of the VMware blogosphere towards Nutanix just doesn't make sense. Nutanix still ships VMware licenses, and it was VMware that decided to step on Nutanix's turf with VSAN, not the other way around. If, however, VMware has been planning a complete converged play all along, a lot of things start making a lot more sense.
Read more from Trevor Pott.
All the rumors point to MARVIN being the same sort of hardware/software combination that is the Nutanix solution. MARVIN seems to be vSphere and VSAN on top of a small, dense blade platform. The talk is of four blades in 2U with a couple of dozen disks, which matches the Supermicro 2U Twin2 range of servers used by Nutanix. I've always thought that the hardware Nutanix uses would make a great VSAN platform, so this would be a logical choice. Nutanix uses its own file system on top of vSphere to deliver shared storage. If VMware releases a product that looks and works a lot like Nutanix, then their recent behavior toward Nutanix makes more sense.
Is VMware about to go to war with its partners? I guess we should say is it about to go to war with more of its partners. VMware has vCloud Hybrid Service (vCHS), [which] is a direct competitor with other public cloud providers that offer vSphere-based services.
We shouldn't be surprised if VMware offers a hyper-converged platform that competes with its partners like Nutanix, SympliVity and a few other startups. VMware may be looking at ways to generate more revenue as they approach virtualization saturation in the North American market. In addition, tightly coupling the software and hardware simplifies support and allows focus on developing product features. Supporting hundreds of server models is expensive for VMware.
One interesting aspect of the image we have for MARVIN is that it looks to me like Google's green Android overlaid with the Gmail icon. I wonder if this is just pre-production internal marketing with some humor, or should we read more into it? After all, VMware got up on stage with Google to show DaaS access from a Chromebook. Are VMware and Google getting cozy? Should we also expect to see a Google-branded DaaS offering to compete with Amazon's Workspaces DaaS? Maybe we will see a version of vCenter that runs on a public cloud platform to manage your on-premises vSphere servers. This would be a way to extend the VMware Go products aimed at small business customers.
My concern is that MARVIN will be a distraction from VMware's core business. Just as SpringSource didn't fit with vSphere, MARVIN may be a product that confuses the customers until it's spun out. On the other hand, buying vSphere is often tightly coupled with buying server hardware.
Read more from Alastair Cooke.