Interview: An insider's view on storage

  • Dan Neel (Computerworld)
  • 12 June, 2002 09:52

With his focus on storage, the datacenter, and distributed computing investments, Jo Tango, general partner with venture capital firm Highland Capital Partners Inc., sees IT from a unique vantage point. The industry insider sat down to examine a few current technology trends and dispel a few industry myths along the way.

Q: From a venture capital point of view, are enterprise storage customers having their needs met by vendors?It all depends on what type of environment they're facing, and I group these environments into two groups of people. One [group is] transaction-intensive environments, [with] a lot of block data. The others are sort of more a file data driven. And the reason that distinction is important is that the folks who are in transaction-intensive environments are sometimes satisfied, never happy. The reason being is that they have huge, 24-by-seven needs -- think about trading for Goldman Saks, think about derivatives for Morgan Stanley. So what they do is a lot of them have in-house capabilities, because a lot of times what the end-user wants from the vendor, the vendor either does not have or [it] is almost in the vendor's economic disinterest to provide. For example, completely open storage software across everybody's subsystems. As a vendor, you have no upside to release all of your APIs. So a lot of financial institutions in particular, say, "I can't wait. I need to build these features myself, so I'll do it in-house." And with a lot of the guys who are moving more file data vs. block data, they're going to be happier, frankly, because their needs are not going to be as complicated. And so a more mass-market solution from a vendor will apply.

Q: Certain storage vendors have recently charged that storage interoperability is a myth and that users buy their storage from their server vendor. Is that true?I think for the most part it is true, by decision maker. The challenge is that there are multiple decision makers in an enterprise. For example, one of my contacts works for large datacenters in Fidelity. They did an audit, they thought that they had a couple of thousand servers, [and in fact] they had 7,000. They thought they had a thousand desktops, they had 32,000. Meaning, that everybody [who] had a budget was buying. So, awkwardly enough, what you find in a lot of enterprises is that one decision maker is choosing one vendor, but the man or woman 50 miles away is choosing somebody else.

Q: Is there a quick fix for such a problem?It is a management problem. However, again, I think different decision makers have different data requirements and therefore storage requirements, because they're supporting such different business processes. Each department has a different tolerance for latency, and therefore different technologies are required to support them. The IT users are making trade-offs comparing price and performance, but every department's graph is different, everyone's trade-off forms are different in the organization depending on what they're required to support. There isn't a homogenous vendor because customers' needs will never be homogenous.

Q: That's a nice way to put it. Can outsourcing, such as the recent IBM/American Express utility computing deal, help companies get a grip on uncontrollable procurement of technology?People love that idea. I know a guy who runs the datacenters at Goldman.

He says [in] his ideal world ... he'd [prefer] buying cheap hardware with software, and if something breaks a vendor is alerted automatically and he or she shows up and fixes it without his ever having noticed. So, currently put, "one throat to choke" -- they want to know whom to yell at. So yes, everybody wants to outsource because data consumption at their companies [is] doubling every nine to 15 months. But conversely, the numbers of computer science graduates in the U.S. continues to go down.

Q: As investors, does this mean Highland Capital is looking for more startups with outsourcing models?We actually don't use that framework at all. What we do is we back people who know the customer pain point intimately. Either because they've been selling to the customer for awhile or they've been customers themselves and they know how to build products. Then we combine them or partner them with a combination of people, depending on what the existing company has; people who know how to build products correctly the first time and people know how to sell service and enterprise correctly. So, that's our approach. Get into a more solutions approach vs. a point solution approach.

Q: You see a lot of new, ambitious startup companies, but do you see a new killer app coming anytime soon?I think the next killer app is video-on-demand.

Q: The billing and transaction processing companies tell me that too.They're right. For example, as cable operators go digital and you can order any movie at any time by the drop, is pretty huge and it's coming. What does that mean for IT and for storage? That means that you need to come up with some pretty interesting storage devices and caching plays at the edge of the network. For example, you'll need to have a bunch of movies or movie clips stored at the edge so when the consumer clicks on "Buy" that movie starts right away, and at the same time you're pinging a core server to start streaming some of that data towards the edge. So there's something, sort of buffering or caching technology that will need to happen, because having copies of full-blown movies around a network, that's a lot of storage.

Q: You mentioned caching as a way to accelerate video-on-demand, but what about bandwidth. I hear that mentioned less and less.Bandwidth is at a different point, which is, if you'll look at access vs. the backbone there's a glut of access and there's a monstrous glut along the backbone. The utilization rates are incredibly low. So I don't think that's the bottleneck. That's the reason why [companies] aren't buying a lot of equipment -- they put a lot of stuff in [and] are waiting for the pipes to fill up. ... I think the bottleneck right now is storage at the edge and technologies that support those types of functions.

Q: Another word I hear less and less when talking about storage is "security." Vendors offer ways to have it, but I get the sense they no longer feel it's a vendor problem, beyond making security tools available.I think security has been a topic du jour, which has quickly faded out after Sept. 11. I think the challenge is [that] security means everything and nothing. When you say security, what do you mean exactly? And that's witnessed by the fact that at most of my enterprise contacts there isn't one security decision maker with a budget because security is sort of a little bit of everybody's job.

Q: Here again we have the dilemma of uncoordinated technology procurement?Exactly. For me, it's always about following the money. And [whether] the enterprise, on average, has been able to articulate and put forward a decision maker and empower him or her with a budget. Now that tells them that there's something interesting going on. But ... I don't think security is a fad. I think it is being addressed. I think security is always keeping up and delivering a better mousetrap. For every defense we put up, the hackers always find a way to get around it. It's a constant, like national security, it's a self-conscious luxury. That's why I think it's such an attractive investment sector. I think what's unclear to me is how do the security startups quickly penetrate an enterprise when often times it's not clear who they should call to get the business.

Q: What's your advice to storage buys looking for maximum ROI?Network your storage. The economics of going from NAS [network attached storage] to SAN [storage area network] to NAS are just too compelling to ignore.

Q: What's the choice? Fibre or IP?It depends. I think if you're a large datacenter, Fibre Channel works. I think iSCSI will be appropriate for the small to medium-size businesses; it's easier to learn. But if you're a block data-intensive environment, that's just made for Fibre Channel. The fact of the matter is the large datacenters have installed it, like it, and it works. But again, I think the question I always ask is not should the reader do x vs. y, it's more what's your business that you're supporting, where are the applications, and therefore what storage do you need to back all that up?

Q: Are you optimistic for a near-term rebound for the IT economy?Cautiously optimistic. Here are the leading indicators I look for: People will invest only if they accept that they have profits, and IT is unfortunately an indicator because IT will have budgets only when they have visibility and people have visibility only when price points stabilize. What I take comfort in is that in Q4 vs. Q1 ratings, the price point went down only 5 percent vs. 40 [percent] to 50 percent. ... We're almost at a stabilizing price point. That tells me that okay, demand and supply are finally being matched up in the market -- the prices are stabilizing, right? So price stabilization is what I'm looking for because then people will start behaving more rationally.