What is the future of cloud storage? A conversation with Jeff Treuhaft, CEO and co-founder of Zetta
By Juergen Urbanski
Q: Jeff, what percentage of enterprise IT demand for storage will be fulfilled by service provider (i.e., external) clouds in 3-5 years? What drives that?
A: In 2014, 80% of new storage capacity and 50% of spend may go to the cloud, driven by the complexity crisis you laid out. The opex for sustaining IT is driving the majority of TCO, even though disk and array prices continue to decline. Most IT organizations will lack the time, budget or experience that will be required to successfully fulfill their own storage needs. Against that, we can deliver our service to customers at one fifth to one tenth the TCO of equivalent NAS-class enterprise storage arrays, assuming street pricing and a 3-year depreciation cycle.
Q: What percentage of enterprise IT demand for storage will be fulfilled by enterprise (i.e., private) clouds in 3-5 years?
A: Longer term we expect to see very few private clouds in the general enterprise space, except for government and perhaps some extreme high performance computing situations. I doubt the average enterprise customer will be able to successfully buy, configure, deploy, manage and grow a private storage cloud. If they couldn’t keep up with data growth in the legacy world, in spite of attempts to standardize on fewer vendor products and interfaces and more storage monitoring and management, what makes anyone think they will be able to handle more hardware, more variable performance components, more monitoring that would be typical of a private cloud?
Q: TechAlpha believes enterprise clouds will play a more prominent role in the large enterprise than external clouds. Where do you see most demand today for infrastructure as a service today?
A: Storage is the one area within IaaS that is ripe for the fastest adoption. Storage access interfaces are already standardized (SAMBA/CIFS, NFS, iSCSI, etc.), application compatibility is defined
by POSIX, and virtualization (aka thin provisioning) is something most storage/IT administrators are already using today.
We are seeing demand across multiple use cases simultaneously. Roughly in order of descending performance and aggregate feature requirements, these are:
1. Business Continuity
2. Data Warehousing
3. File share/Home Directories
5. Data Migration
8. Active Archive
Q: What does that imply for the longer-term level of profitability for storage clouds? Will scale matter more than customer intimacy or service innovation?
A: Innovation that drives automation and operational efficiency matters much more than purely scale. The original crop of storage service providers failed because their business model and underlying
technology were too costly. Storage Networks for instance raised $400m+ and grew to more than $100m in run rate revenue. What makes the economics of our business work as we grow is the combination of low capex costs, leveraging commodity off the shelf systems, with our unique file system that was purpose built to enable us to sell, operate and scale at low cost.
Q: As storage demand goes to cloud services, is there any way for legacy storage OEMs to serve that market profitably, given their business model and technical constraints (no scalability, poor automation, legacy design point)?
A: Commercially, their component costs (using custom ASICs, etc.) and packaging are cost prohibitive when compared to the commodity off-the-shelf model.
Technically, most file systems and array architectures were built for customer on-premise usage. Features like native multi-tenancy (i.e., not thin provisioning) are non-existent. Today’s array architectures are largely “blocking based” in terms of the view/data path from the application interface to the disks themselves. No amount of clustering, virtualization or other attempts can belie the fact that the costs and complexities of those architectures will limit any service built upon them.
Culturally, building and operating an enterprise class storage service is a completely different organizational design point than trying to compile and ship hardware or software products. Their ethos, history, talent, sales model, etc. are all geared toward selling and licensing product to customers.
Q: What go-to-market approach works for you? Is this an infrastructure buying decision, or is it driven by the platform or application buying decisions which have traditionally driven storage
A: All enterprise storage clouds will have to be app aware to survive. Most of those apps only really work where there is a POSIX compliant, secure, protected storage system that can be accessed by native protocols and delivers a strong consistency model. In enterprise class cloud storage there is also no such thing as “storage-only without compute” if done properly. Zetta for instance has a
significant amount of general purpose compute already provisioned as part of our storage system.
We’ve already signed quite a few partners in the VAR and SI space. These are usually folks who are traditionally selling storage arrays and networking equipment or asked to develop and execute technology transition strategies and projects. Longer term, significant value will be driven by those that can help the ISV community really crack automation and optimization of storage.