Traditionally, compute, storage and networking capacity have been purchased as separate resources, with largely independent management structures. Indeed, in most organizations for example, the networking and storage teams have different reporting lines and are measured on achieving different goals. Again, by way of example, networking architectures are highly flexible and designed to accommodate new users quickly. Networking professionals often need to reconfigure the network to support new business growth. Storage on the other hand, particularly block-based SAN storage is a different animal. Usually once the SAN is hardened, storage admins don’t like to mess with the infrastructure and make changes to the system unless absolutely necessary. SAN managers are intensely focused on data reliability and integrity whereas in networking, if data is dropped it can be re-submitted without any major disruption to the business.
This organizational dynamic is one of the most interesting confronting data center professionals looking to acquire so-called converged stacks. Converged stacks refer to the grouping of server, storage and networking hardware, software and infrastructure management into a single logical resource that is virtualized. The idea is that by installing a logical piece of virtual infrastructure that satisfies the need for all three areas, applications can be served in a horizontal manner, lowering complexity and costs while at the same time speeding deployment and change. In theory and in practice this can better serve business lines and support the move to IT-as-a-Service and cloud computing.
This move is a big bet being made by the whales in the IT industry, including HP, EMC and Cisco (through the VCE coalition) and Oracle which is also converging infrastructure but in a different manner. Smaller players such as NetApp are participating as well while IBM has been notably quiet in the converged infrastructure space. The approaches of HP, VCE and Oracle have varying degrees of value and risk to customers that can be summarized as follows:
- The primary value proposition of all these systems is simplification of management, lower costs and speed to deployment; while at the same time reducing risk by having an integrated solution.
- HP is in my view the most open allowing more combinations of its infrastructure components than other vendors.
- Oracle is the most closed and the most integrated throughout the hardware, software and applications stack, choosing to not only engineer hardware components (server, storage, networking) but also database and other Oracle software.
- VCE is somewhere in the middle, choosing a partnership model that involves EMC, VMware, Cisco and Intel, as well as selective partners in the ecosystem.
Is this a Winner-takes-all Strategy?
In some respects, yes – the vendor that can get customers to buy into its stack will gain the benefits of securing key pieces of the stack which it may not currently own. For example, Cisco is a new entrant into the server space and by using its networking leadership in combination with its EMC and VMware partnership, it can pick up server share by default when users choose to buy Vblocks (VCE’s converged stack product). The challenge for VCE however is that customers may be happy with their existing server vendor and might not want to give up that relationship. In the case of Oracle this is especially true as it’s an all-Oracle, all the time approach– like the iTunes stack for the enterprise. But the prize is big in that if VCE or Oracle win the converged business, they are likely stealing share from an incumbent server, storage or networking player.
HP has a different approach. Because it allows customers to mix and match components it can compete on the basis of converged infrastructure while at the same time not forcing customers into a single stack approach. Clearly lock-in with the HP strategy is less than with the other approaches.
I asked Dave Donatelli about this dynamic recently on the Cube at HP Discover. Here’s what he told us:
- Competitors’ stacks are mainly narrow
- Competitive stacks are mostly “any way you like it as long as it’s only my way”
- HP is horizontal and covers more apps
- HP is building to open standards to allow both HP and non-HP infrastructure to be included
- HP’s approach has two main advantages including showing a way to the cloud that includes legacy infrastructure and at the same time doesn’t lock customers in
Here’s a two-minute clip of Donatelli comparing HP’s approach to the competition.
Squinting Through the Marketing Hype
Which Approach is the Best?
Of course every vendor will tell you their approach is the best way forward.
The ultimate answer to the question depends on your goals. Oracle has a strong case that its stack is the most complete with Oracle engineered through the entire system, including, the hardware, hypervisor, middleware, database and application software via its Fusion architecture. VCE can claim that though its partnership strategy, huge investments in the joint venture (VCE has 3,000 employees today) and its single SKU approach, that it has both engineered a complete solution and checked the openness box by allowing any software to run on the system. HP can claim the most open architecture because it will allow any hypervisor, any systems management software, any application software and the ability to mix and match any server, storage and networking into its systems; while at the same time engineering a complete stack solution that’s all-HP for those customers that want such an approach.
What’s happening in the industry in my view is that there have been several moves made on the chessboard that are forcing suppliers to react and respond. Specifically:
- Cisco has VMware envy. Cisco missed out on buying VMware and needs a growth strategy that includes virtualization. Cisco’s move into servers is predicated on partnerships with most importantly VMware and secondarily EMC. Intel is a no-brainer because clearly Cisco is not going to design proprietary servers. EMC is fundamental to Cisco because it owns VMware but the reality is that Cisco could care less about who wins the storage business which is why it has strong ties to NetApp as well as EMC.
- Oracle is trying to take over the IT world. Since Ellison started writing more checks than code to acquire software and hardware companies, the game plan in the IT industry has changed. It’s all about converging stacks and engineering hardware and software together. This creates the ultimate lock-in strategy which is always more profitable for those who can achieve it. The question is will the allure of Oracle’s integration offset the inherent risks of getting locked into Oracle, the most expensive approach in the market with very onerous contracting terms.
- HP doesn’t own the hypervisor and it doesn’t own database and applications software. So it has chosen to feature open-ness as a benefit. It is true that HP is more horizontal because it supports more hypervisors (e.g. Hyper-V), more OSes and more choice…so naturally it’s infrastructure can play in more places. The question for HP is can it compete on the basis of best-of-breed with Oracle and VCE who both offer less choice (e.g. only one hypervisor) with converged infrastructure but can argue higher degrees of engineering.
My take on this last point is yes. HP has an all-HP strategy and it supports alternatives. Both VCE and Oracle by the way would say the same thing – i.e. VCE would say if you want to by EMC storage with Hyper-V and Dell servers, go for it…what’s the difference from that and HP’s approach?
The difference technically is nuance. All these companies make good products and they stand behind them. Depending on your skill sets and technical preference you can make a case for any of them. The real difference is positioning and marketing…which often makes all the difference in the world.
The bottom line: HP, VCE and Oracle are leading the way in integrated/converged stacks for the data center. HP has tons of upside potential to gain in networking and storage and is the most open. Because its coming from a lower margin structure, HP will be a value pricing play. Oracle is going for a monopoly-like, high profit strategy with a closed approach at a price premium; while VCE is creating new partnership models and is talking open but its profits will be made on its own stack.
ServicesAngle: The services dimension here tends to get lost in all the technical hype. The reality is that traditionally services companies have been called upon to do much of the integration around making compute, storage and networking resources application ready. In the future, increasingly organizations will buy pre-integrated packages forcing services to move up the value chain. The way for services organizations to leverage this trend is to market and in some cases build cloud-ready infrastructure for their clients. What does this mean? It means providing services to accelerate converged infrastructure adoption and/or directly using converged infrastructure as part of a cloud offering that is sold on a rental basis to clients.