At its Network Computing event on May 2 in Washington, D.C., Sun Microsystems unveiled a host of new storage offerings and upcoming products poised to launch the systems management company into the upper echelon of the storage provider elite. Mark Canepa, executive vice president of the Data Management Group at Sun, talked to Senior Writer Brian Fonseca about Suns unique network capabilities to address evolving storage requirements, the future of tape-based systems and how to perform better data housekeeping.
What advantage does Sun have toward building better storage systems and information management tools over other large systems vendors and pure-play storage providers?
Weve tied [storage] directly in place with Sun Microsystems. Theres a computational side to the network that is all about Opteron and SPARC and Solaris and middleware, basically providing an infrastructure layer for applications to attach themselves to so they can be developed at a much higher level than writing to an operating system. So if you now look at that and you start to take that application environment to its logical conclusion, what a few years ago in the data center was a few or few dozen fairly well-behaved servers—each running their own application environment—now that thing has become 100,000 instances of "containers," heterogeneous perhaps, and ultimately a grid.
So if youre computing environment goes from 10 to 100,000 [containers], what does your data environment look like in order to be able support that? Youre certainly not going to be putting 100,000 separate physical storage entities out there for each one of those containers to have. We have to be able to express onto [the] network a set of protocols and APIs that we call data services to allow 100,000 containers to think they have 100,000 storage environments, but the reality is you better implement those on a few dozen physical things, or you will never make it happen.
So in terms of ROI, what type of value proposition does that enable?
The value proposition is simple. Provide a set of data services to network that these applications can use so that the data that they need can be trusted, available and has the right value. We believe were uniquely positioned to implement that strategy. Because if you go underneath that data grid and data environment and you look at what you need to implement that strategy, not only do you need disk and tape, you need to be able to provide a layer of software that sits between the physical environment so you can create the data services. So you need an Opteron system, you need middleware, you need system-level know-how and Sun is one of the few companies that have all the ingredients to pull this off.
Does Suns approach toward simplifying and optimizing storage utilization start from the server side on out or vice versa?
It ends up being a little bit of both. We start at the network. If you start with the network and the transform mechanism is either an Infiniband network or Fibre Channel, then you see what are the APIs and protocols that people want to talk to [each] another. What do the applications want to see when they look at data? There are all sorts of much more interesting and higher-level interfaces that applications really would like to talk to storage; the biggest one is XML. Why do you want to use NFS [Network File System] when you can say heres an XML object? Why not give the storage content knowledge?
[For instance,] heres an API that lets me delegate search. Search is going to become a huge interface with storage. What youre going to see is over time, storage boxes are going to be enabled to deliver much more intelligent things. That intelligence is going to require the worlds best operating system as an environment to write those storage-intensive applications. We have it. I can reach into 3 million Java developers and be able to build storage applets that sit inside my storage box.
XML is not something you traditionally hear about from storage companies.
Well, of course, thats because were a systems company. You talk to a storage company, the words you hear are RAID 6 and data replication. Thats a little like saying you better put plumbing in your basement. The first floor and second floor are marble bathrooms, which are search [and] XML—theyre these higher-level value layers. When EMC talks to you about the plumbing in the basement, were past that. Where is [EMCs] Centerras ability to do XML? Wheres Centerras ability to do anything open standards? At best they push it to a proprietary protocol.