A few weeks ago at GigaOM’s Structure event, we hosted a Mapping Session to sketch out the implications of Infrastructure as a Service’s impact on the enterprise data center value chain. We asked about 40 GigaOM Pro subscribers to join in an interactive discussion with a panel of analysts to assess how far and how fast IaaS would commoditize storage and networking hardware, and the management software that runs it.
Our Mapping Sessions often comprise the first stage in our process for developing research that we present as a Sector RoadMap report or other ecosystem analysis. In a Mapping Session we aim to tap the collective wisdom of our analysts and other thought leaders in the market sector we are researching, to identify the most disruptive trends and their relative importance in shaping a market or industry over the next 12-24 months. We also want to examine which companies are in the best and worst position to ride or drive these trends – we call them “Disruption Vectors” – to gains in market share and revenue.
In this Mapping Session, “The IaaS endgame: commoditizing the data center,” participants decided that computer hardware technology disruptions were so far advanced as to no longer be Disruption Vectors that smart companies could leverage. The collective intelligence present at the session determined that advances in CPU core density, or the impact of Facebook’s open compute standards, were pretty well understood and not big opportunities for vendors. Server gross margins have been commoditized into the single digits, but for the moment shared storage in the form of NAS and SANs are still in the 50 percent to 60 percent range. Routing and switching hardware hovers near 70 percent.
Session participants settled on three critical Disruption Vectors:
Storage commoditization. Plunging flash memory costs plus virtualization software will drive the margins out of big storage first. Network-attached storage (like NetApp) is already becoming a commodity server running as the storage controller connected to an array of disks. However, the storage software isn’t yet able to span controller/arrays from multiple vendors. Storage area networks are on a slower commoditization pace. SANs are typically engineered for extreme performance and fault tolerance.
Network switch commoditization and SDN. Networks from Cisco, Juniper, Arista, and Nicira are heading towards commoditization almost as rapidly as storage, with software-defined networks running switching and routing intelligence in software atop commodity servers and storage is also on the path to commoditization. Though controller software can run on commodity servers, the admin tools don’t interoperate that well. That implies single-vendor networks for a while, where the vendor has pricing power, even if it’s built on commodity hardware.
Open source impact. Many critical cloud-computing technologies had open source origins. Panelists and audience participants expect the open source world will continue to power innovation, and likely spawn management software and new tools for things like workload migration. However, open source is still no defense against supplier lock-in.
Three less influential Disruption Vectors were:
Management/orchestration tools. More means to the end of data center commoditization via IaaS. With the interoperability potential to break down some of the aforementioned lock-in and pricing control.
Packaged software business models. Most IT departments don’t have the experience to create and manage a software business the way packaged software or IaaS companies do. That’s a potential opportunity for systems integrators and other service providers.
Power consumption. Earlier, during the onstage Structure sessions, Zynga’s CIO Debra Chrapaty said electric power consumption could account for 40 percent of the cost of running a big data center. Mapping Session participants said that figure often goes to 75 percent. Consensus was that low-power hardware and management software represented a good opportunity for differentiation and some hope of margin preservation.
Implications for data center vendors
Service provider-scale data centers may have highly optimized modules at a completely different scale than those suitable for enterprise data centers. 40-foot containers may come in several options. One may be configured with the highest performance, highest availability SAN storage. Another container may be configured to have the highest performance routing in order to introduce the minimal possible latency into a mobile app.
Custom collections of modules are generally low-margin, highly commoditized products that lend themselves to very specific use cases. Highly integrated containers can be assembled to add huge amounts of specific types of capacity in order to stand up new data centers very quickly. These containers are likely to support higher gross margins than their off-the-shelf rack or blade brethren. Specialized module configurations will emerge for specific types of computing tasks like transaction processing or high-performance computing in support of scientific or graphical applications.
EMC wants to ensure if damage is done, it does it to itself. It announced a reorganization of its PaaS and SaaS businesses shortly after our session, partly for focus, partly for a shot at new opportunities, and partly to separate businesses with fundamentally different pricing models that drive their valuations.
IaaS providers are moving up into the PaaS space, and PaaS players have their eyes on the applications layer provided by SaaS. But those are tales for another set of Sector RoadMaps.
- Enterprises will be able to share or rent out their own data center in IaaS-enabled marketplaces, but that will play out beyond the 36-month horizon.
- Session participants decided IaaS data center security was less a disruptive force than an adoption barrier that presented modest differentiation opportunities for would-be cloud infrastructure suppliers.
- While cloud-based data centers enabled by IaaS are viable in countries with robust high-speed network availability, that is not the case in many regions of the world. Data center commoditization will thus develop at different rates, depending on regional infrastructure maturity.
We welcome your feedback on these disruptive trends. Have we missed or mis-emphasized anything that you believe will be key to shaping data center commoditization over the next two years? Continue the discussion by leaving a comment below.
Mapping session panelists