NetApp’s June 2013 release of ONTAP 8.2 is the next incremental step on the company’s journey to re-form its product portfolio for the new era of IT-as-a-Service and cloud. Clustered Data ONTAP is a radically new approach for NetApp’s storage architecture that the company has slowly rolled out over a period of several years. The latest release adds significant enhancements primarily focused on non-disruptive operations (NDO) and improved scalability, while maintaining the historical efficiency benefits inherent to WAFL, the company’s unique file system.
In our view, Data ONTAP 8.2 will drive a significant new wave of adoption within NetApp’s customer base, pushing the 8.x uptake to over two-thirds of the company’s installed base. In addition, we believe the non-disruptive benefits of ONTAP 8.2, combined with its leading OnCommand management suite, will allow the company to reach critical mass with its clustering technology, setting NetApp up for two key objectives:
- Maintaining its loyal customer base, and,
- Making a run at higher end markets.
Moreover, in our view, NetApp is well positioned to compete in the so-called software-defined storage space due to its inherently virtualized architecture, granular storage service offerings and propensity toward industry collaboration. Its position as the last large truly independent storage ‘pure play’ will make NetApp an attractive partnering candidate for many ISVs.
Betting the Mortgage on Clustered ONTAP
In our recent piece on the future of NetApp, Wikibon shared its view of the importance of clustering to NetApp and its customers and the challenges of making clustered systems work at scale. In this narrative we explained that while customers love NetApp’s products, at scale they became difficult and complex to manage. One factor we didn’t discuss was the challenge that for years now NetApp has had to maintain multiple code bases -- its legacy ONTAP and its “C-Mode” (clustered).
Maintaining two code bases has always been a challenge for IT practitioners and technology vendors. We’ve seen this many times in the industry. Two prominent examples include migrating custom Cobol code off of mainframes and Oracle’s migration to Fusion Applications, which took upwards of seven years. The difficulty of migrations for custom code (i.e. non-COTS software) is well understood by IT practitioners and technologists and is often compared to changing out an engine in mid flight. Specifically, a hard and fast rule of migrations for custom code bases is that unless the legacy code is frozen, the migration will be painful, risky and take forever (figuratively speaking).
In NetApp’s case (as in Oracle’s) it couldn’t freeze the legacy code in the early days of development because it would have lost ground to the competition. NetApp needed to continuously add features to its historical code base to maintain innovation or it would lose significant market-share. The problem with this approach is the new code base (Clustered) had to be updated to keep pace with the legacy code base’s feature sets while at the same time making the newer technology stable. So the task is endless until the new code base is far enough along that the legacy code can be frozen.
NetApp is well past that point now, and all its innovation can be poured into its clustered code base. The key for NetApp is adoption, which will further cement its customer loyalty, protect its base, and allow it to gain share into higher end spaces.
Addressing Management Complexity at Scale
NetApp’s main messaging for 8.2 focuses on three key areas:
- Non-disruptive operations,
- Seamless scalability,
- Proven storage and operational efficiency.
Of these, we believe non-disruptive ops will have the greatest near-term economic impact for NetApp’s customers. Specifically, [Wikibon community research shows] that array migrations can comprise upwards of 30% of total costs over the life of an array. As we’ve pointed out many times, a key advantage of NetApp’s architecture is simplicity and the ease with which resources can be moved around the system. But as the number of boxes on a customer’s floor grows, manual movement and array migrations become more difficult, because customers have to incur planned downtime to make changes. With ONTAP 8.2, NetApp is promising true non-disruptive migrations and general operations (i.e. moving resources won’t incur downtime).
Regarding so-called “seamless scalability”, it’s unclear exactly what this means. Observers could infer full automation, but we don’t believe that is the case with 8.2. In other words, making changes within NetApp’s clustered architecture is easy and can be done without taking the system down; however our understanding is changes are still manual. Because we believe NetApp’s clustering technology is aimed at large-scale installations (over time) this capability will need to evolve to support higher degrees of automation than exist today. This is critical, because as customers add nodes and workload complexity grows, changes will have to be made to balance system performance. To the extent these changes can be fully automated, it will give NetApp significant competitive advantage.
Regarding “proven storage and operational efficiency” we have no argument with this statement. To the extent that NetApp addresses planned downtime in 8.2 – which it does – the inherent benefits of WAFL will be seen by customers.
OnCommand: The Supporting Actor
A key capability not mentioned in NetApp’s 8.2. press release is OnCommand. This is NetApp’s management suite that was developed by combining internal NetApp management software with external acquisitions including Onaro and Akorri. As we’ve indicated before, OnCommand is a key component of NetApp’s strategy, as it enables many of the benefits of clustering to be realized. In addition, because we feel the future of ONTAP is largely about automation, we believe OnCommand will play a critical role.
As shown in Figure 1, OnCommand comprises six modules that span low-level device management through a more “helicopter” level view of installed resources. OnCommand is considered a best-of-breed storage resource management offering and competes with the likes of EMC’s management suite and IBM’s Tivoli software.
Importantly, NetApp has put considerable effort to align its OnCommand portfolio with Clustered Data ONTAP; making it somewhat surprising that OnCommand got barely a mention in NetApp’s 8.2 press release. Specifically, the entire OnCommand portfolio has support for clustering today.
We’ve often stated that OnCommand enables the key functionalities of clustered ONTAP, specifically the three pieces that NetApp mentioned in its press release - non-disruptive operations, seamless scalability and proven efficiency. OnCommand in many respects is a key to adoption of clustering because it gives customers visibility on how to manage complex scale out environments by identifying problems, automating workflows and connecting up into external management systems. In particular, NetApp’s WFA product line is critical to clustered ONTAP adoption, particularly in cloud-like environments, because it allows customers to get rid of scripts at scale. Notably, subsequent to the Akorri acquisition, NetApp has rationalized the OnCommand portfolio to provide more focus on deeper NetApp integration (e.g. Balance is now solely focused on NetApp storage), leaving the Insight product to focus on heterogeneous capabilities.
The bottom line is that as the world moves to software-defined environments, instrumenting the installation becomes critical, and products like OnCommand enable that line-of-sight for IT practitioners.
Software (Defined) Eats the Storage World
We’ve often stated that software-led infrastructures will evolve as follows:
- Underlying hardware will be virtualized,
- Both performance and capacity will be abstracted and available through API calls,
- Open and complete object-based APIs will evolve and mature to enable the automation of highly granular storage services,
- Higher-level frameworks (e.g. OpenStack) will be able to access and orchestrate the delivery of these services through open APIs,
- Open source and collaboration will play an increasingly vital role within software-led architectures.
NetApp has always had a highly virtualized backend, satisfying point #1 above. In recent discussions with NetApp CTO Jay Kidd, we specifically asked:
- Can a customer abstract both capacity and performance?
- Can a customer for example provision say 50GB of storage and 1,000 IOPs?
- Can this be done through an API call and can this policy be adjusted on the fly through software without scripts?
The answer to all these questions was ‘yes.’
We then asked Jay Kidd to describe the completeness of the API with respect to the degree of automation of storage services that could be specifically accessed through OpenStack (as an example) to fully exploit these services. The answer was somewhat measured, but it was clear this is a high priority for NetApp.
Regarding support for OpenStack, we did a search on GitHub to identify the most prolific contributors to Grizzly/Cinder. NetApp showed up on the hacker list but could do more as the likes of IBM, HP and, interestingly, SolidFire, appear to be somewhat outpacing NetApp in this regard. Nonetheless, NetApp is clearly contributing to the Open Source movement and notably appears to be outpacing EMC in terms of contributions.
Action Item: On balance we believe 8.2 is a major milestone for NetApp and will accelerate Clustered ONTAP in the marketplace. While practitioners and NetApp customers should still use caution with regard to clustered approaches, generally we believe scale out will win in the market. The non-disruptive capabilities of 8.2 represent a tipping point for NetApp in our view and dramatically reduce customer risk.