This past week VMware made their new virtual networking product, VMware NSX, the star of VMworld, their premier North America event. But as VMware flexes its muscle at VMworld both excitement and apprehension surround the announcements and their plans for future leadership in network virtualization. We believe that those who claim VMware delivered a solid product that takes the game, set and match might be a bit ahead of themselves.
VMware is setting its sights squarely on the network now as they own the lion’s share of the compute virtualization market. Network virtualization is the hottest topic in networking today and the first real change to the staid networking world for more than 20 years. For the first time there is a true disruption that challenges the status quo with the Software Defined Network (SDN) starting to come into discussions with customers as they try to tackle their increasingly complex networking issues.
VMware spent $1.2B on the acquisition of Nicira and now is their opportunity to begin to capitalize on that purchase. But we do not anticipate that NSX will take over the network virtualization space like ESX did for server virtualization. 10 years ago server virtualization was a much easier value proposition for VMware because:
- It all sat inside of one domain (servers)
- It could be piloted in small projects and eventually scaled out as expertise grew and ROI was demonstrated
- There were significant physical constraints that were limiting the growth of server deployments (power, cooling, space) so there was a compelling reason to find a solution
With SDN, the final solution crosses multiple domains, deployment needs to be more widespread in order to truly gain ROI and while there is a lot of pain around networking today, none of those pains is a “show stopper” that is preventing growth.
Faced with these challenges, network virtualization will be a more difficult task than server virtualization, so VMware will have a more difficult challenge here. In an environment where VMware does not necessarily have the market to themselves (as they essentially did with server virtualization,) selling NSX will be a harder task.
What Is NSX?
VMware NSX is a new product that delivers network virtualization, or, more simply, a virtualized network overlay that sits on top of a physical network infrastructure and allows administrators to manage network functionality and provisioning virtually, from a console, without having to touch physical devices or pull cables. This product is the realization of the Nicira acquisition and it helps jump start VMware’s attack on network virtualization.
Not fully available until the fourth quarter of this year, NSX promises to be the “glue between a layer of virtualized servers and a cloud infrastructure.”
With the announcement, VMware was showing off two customers, eBay and Citi who were using NSX now. eBay was already a customer of Nicira long before the acquisition and Citi is a long-standing VMware customer, so there were no surprises here. Where there were surprises however, was in the partner space. VMware is partnering with a number of companies that many may even view as competitors. Is this a new era of cooperation or is this a strategy of surrounding and capturing? Time will tell.
NSX is not an SDN controller, but instead an SDN platform, so it has the ability to work with a range of SDN controllers including those of HP and Juniper. This means that theoretically any SDN controller could work with NSX, but customers would be crazy to use one unless it is explicitly supported, and to date both VMware and Cisco are mum on actual support for Cisco ONE on NSX. This leaves the biggest question in the SDN market, how will the relationship between VMware and Cisco play out in the long run. In the battle over SDN supremacy the lines will be drawn around these two players and the long term shape of the market will be reflected by how these two work (or don’t) work together.
VMware vs. Cisco
In 2009, Cisco announced their Unified Computing System (UCS), which was a significant shot across the bow of server OEMs. For years these OEMs had worked with Cisco and even helped promote Cisco networking products in certain situations, even though the OEMs also competed with their own networking hardware. With the launch of UCS, VMware was a key partner for Cisco, and the UCS platform focused on virtualization, using features from VMware that other OEMs had not integrated (including a very large, proprietary memory footprint.)
Now, with the launch of NSX, Cisco finds itself in the same position as those server OEMs with a partner encroaching in their space and threatening to capture both revenue and margin from a market that is already established and stable. Four years after the UCS launch, Cisco has less than 5% of the revenue for the x86 server market, an indication that while a new entrant into the market can shake things up, actually pulling business away in a competitive (and conservative) market is more difficult than expected.
In the networking market, Cisco is not only the incumbent but also the most entrenched player, with deep ties throughout the market as well as within customer datacenters. While Cisco has its own problems with SDN, the driving issue today with SDN still revolves around how quickly customers move to this new technology. Today, there is still a concern that while SDN is getting headlines, it might not yet be ready for wide scale deployment yet. That gives both Cisco and VMware time to build out their stories ahead of actually needing to truly fight it out head to head. Today SDN is a battle of words; in the future, however, it will be a battle of customers and revenue. At that point things could really get ugly.
Cisco has the premiere networking hardware position with both an impressive install base as well as a huge reach within data centers. However in the SDN world, there are two viewpoints forming on hardware. One group believes that SDN hardware appliances and SDN enablement of hardware is the future, the other believes that hardware becomes a commodity and that the real money is in software. Cisco is a hardware-centric company and VMware is software-centric, so it is easy to see where these two line up. In a software-centric world, inexpensive white box hardware is utilized, leaving more IT budget on the table for software stacks, which benefits VMware. In a hardware-centric world there may not a significant acquisition cost savings which makes the Cisco hill a bit harder to climb.
Cisco’s biggest asset (beyond its hardware install base) is its channel partner army. These are all companies that have made a lot of money installing and servicing Cisco equipment. In the battle over SDN, these partners are going to need to place their bets on whether SDN is hardware or software-driven. VMware has an opportunity to potentially switch the allegiance of some of these partners who see the writing on the wall for a software-based networking future. However, because the SDN change will happen slowly for the broader market, these partners may be reluctant to pass on lucrative networking hardware revenue today in anticipation of a software-based future state.
Both Cisco and VMware have healthy advertising and marketing budgets. With so much on the line as the networking world moves from physical to virtual, expect a lot of noise from both players. In environments like this it is important for customers to remain objective as both companies will be vying for mind share while pitching two very different views of the future state.
Open Source/Standards or Vertical Integration?
Another logical split in the SDN world revolves around the idea of open source and industry standards versus vertical integration. With open source the customer maintains the direction and control of the platform, integrating the pieces that make sense for their business. This leads to a solution that is more optimized and tuned to the exact customer environment; however this also requires more work on the customers’ side. Contrast that with a vertical solution where a vendor provides a complete turn-key solution, giving customers one-stop-shopping, but at the expense of less flexibility and less customization. Seeing how customers are moving towards SDN to free themselves of inflexible networking environments, the thought of vertical solutions will be a tougher sell.
Aside from the efficiency and flexibility that SDN brings, one of the other big drivers for customers is the reduction in acquisition costs. Today customers are generally tied into companies like Cisco or Juniper for their networking, and this is expensive as well as limiting. The idea of turning to open source, focusing on software and selecting the pieces that best fit your business is appealing, mainly because there is an underlying cost savings that comes along with that strategy.
When you ask customers about their perceptions of VMware or Cisco, you rarely hear someone rave about the cost savings. Typically the response is that while they have very good solutions and there is a high level of trust, but customers are overpaying and believe they are locked into a proprietary stack.
With so much of the focus of SDN on breaking the traditional paradigms of networking, it is a good bet that customers aren’t going to be as interested in breaking the chains of traditional networking, only to find themselves tied back into a high-priced vertical solution that locks them back in to a single vendor. This perception will be a significant hurdle for VMware as they drive for market acceptance of NSX. And other vendors, like Cisco, will face the same challenge.
Integration into Non-VMware Environments
Integrating NSX into existing VMware ESX and vSphere environments will be a relatively straightforward exercise, but what about non-VMware environments or mixed environments? The answer there is a little less clear.
From the early days of x86 virtualization, VMware held a virtual monopoly on the market, but over time they have fallen to less than 60%. Expectations are that with other options such as Hyper-V, KVM and Xen, VMware’s share will continue a gradual downward slope, probably ending up in the 40-50% range as three to four platforms share the server virtualization pie.
While many customers had experimented with multiple hypervisors (44% are said to be running more than one hypervisor in their environment) the market will eventually shake out as most customers standardize on one in order to reduce complexity and cost. While most agree that customers are looking to consolidate, the varying numbers make it difficult to determine where the market ends up at this point.
Consolidation is both a blessing and a curse for VMware. For those that standardize on VMware as their hypervisor, NSX will have a clear advantage. Customers standardizing on VMware are mainly voting down the path of vertical integration (and higher licensing costs) already, so the NSX value proposition works.
Customers making the conscious decision to move away from an all-VMware environment typically fall into one of 3 categories:
- Customers with mainly Linux shops that see KVM or Xen as being better aligned to their Linux environments
- Customers concerned about the high price of licensing. Typically they are utilizing VMware for their most critical servers and utilizing another less expensive hypervisor for their less critical workloads.
- Customers concerned about lock-in and worried that one vendor can have too much control over their IT environment.
While it may be difficult to get all of the research numbers to align, the following universal truth can be gleaned from the data: There will be more than half the market that is either not running VMware or not running VMware exclusively. This makes the pitch for NSX potentially out of sync with the majority of the market. There are some that are so concerned about the cost of VMware licensing that they are willing to take on more operational cost from running two (or more) hypervisors. Clearly, while VMware has good products and is the “Cadillac of virtualization” they have not only gained that moniker from their products but also their prices.
Our belief is that in VMware-only environments (or predominantly VMware environments) NSX stands a chance for high adoption rates, but in environments where VMware is either not being deployed or is not the primary hypervisor, customers will resist NSX, viewing it as an expensive and limiting choice.
There was quite a spirited debate on Twitter about NSX and its impact on the people in the trenches that live and breathe this stuff.
Just as SDN has the ability to disrupt the existing market, it also has the ability to disrupt the people and functional organizations that are tied to it. Back in my sales days, all of our racks came with a common key. One of the more popular rack options was a lock core with a unique key, ostensibly so that the server guys could keep the others out of “their racks.” There have always been silos in IT, but with SDN (and most especially NSX) those silos start to blur. A lot. This can be both a positive and a negative.
When SDN is proposed and the networking team is leading the charge, everyone knows their place. But with NSX, is that direction driven by the server teams? Do server admins need to now become networking experts? Does this mean that networking experts are now at risk? There are a lot of unanswered questions.
Realistically, in the typical IT shop there is a.) more work to go around and b.) not enough resources to get it all done. That does mean that these strategies won’t necessarily trim IT, in most cases (at least in the near term) there will most likely need to be more cross training as organizations begin to flesh out how they will staff and run their functional areas. And while they are looking at this blend of servers and networking, now is probably a good time to look at storage as well. With new strategies like distributed object stores with CEPH there is an opportunity to really rethink more than just networking.
What Does This Mean For NSX?
Although it is not even out on the market yet, there has been a lot of discussion about NSX and the jury is definitely out on the strategy. Here are the key points for the market:
- This clearly turns the SDN market into a three-horse race. The incumbent, Cisco, is riding their market share and mindshare horse. VMware is coming up quickly on the horse called “fully integrated environment.” And everyone else is riding the third horse which includes open source, industry-disrupting startups and some familiar faces (Dell, HP, IBM, etc.)
- The SDN market is looking to break free from the chains of rigid and inflexible physical networking. As the 800-pound gorilla of the server virtualization space, they need to make sure that they don’t try to co-opt the networking virtualization space by claiming de-facto leadership. Customers are already suspicious.
- The announcements at VMworld were clearly pitching a platform and partners, it is critical that VMware not only expands these partnerships, but also helps ensure the profitability of their partners. Part of being in an open environment is ensuring that everyone gets a piece of the pie; that is how long-term ecosystems work.
- IT pros are clearly skeptical, so the proof is in the pudding so to speak. NSX has an opportunity to have a significant change in IT, but only if the people tasked with evaluating and deploying it are part of the solution. There is still much education that needs to happen in order to ensure a fully engaged set of professionals can carry the NSX momentum forward.
Again, we believe that while VMware’s NSX product has some compelling benefits, it is too early to call it a winner in the market. In VMware-centric enterprises it will be a much easier to sell, but in multi-hypervisor environments or environments that still maintain a large number of non-virtualized servers, deploying NSX may not have the same benefits.
John Fruehe is a guest blogger at Moor Insights & Strategy and was previously vice president at NextIO and director at AMD. John also spent nearly 15 years at Dell and Compaq in enterprise product, strategy, and marketing roles. You can find his full biography here.