Tuesday May 15, 2007

Diary of a startup - lessons learned

I drove out to Orange County last week to inspect the startup's new offices. They were pretty impressive, on the fifth floor of a newer highrise. They're definitely planning for future growth -- most of the desks were unoccupied. "Monty," the CIO, showed me around. The hardware is up and running, and it hasn't been moved to their co-location center yet, so I got to see it whirring away. I guess computers haven't been visually impressive since the days of big iron -- today they are just a few slim boxes sitting in a rack that you could wheel into the elevator -- but they do retain an undeniable elegance, a sense that they are the embodiment of an enormous amount of intellectual effort.

Monty and his team have done a terrific job integrating everything in a very short time: HP boxes, Cisco routers, VMware virtualization, Windows operating systems and database, BEA middleware, specialized off-the-shelf mortgage applications. They should be funding their first loan this week, only a few days beyond their original four-month schedule.

There are some key lessons that can be drawn from watching the way this startup came together, lessons both for Sun and for prospective entrepreneurs. Naturally, these are only one man's opinion.

The first is that we are in an era of intense specialization. Monty pulled together both hardware and software from many different vendors, choosing what he saw as best of breed every time. A corollary for a multi-line vendor like Sun is that to appeal to this type of buyer, you have to be among the best in every product line that you hope to sell. If you aren't, you're going to get cherry-picked. Pros like Monty don't set much store by one-stop shopping. The reason is obvious: startups are so inherently risky that their IT crews will do anything they can to mitigate risk, and the best way to do so is to buy what you perceive as the best.

The second lesson is that open source software without a vendor behind it has a huge handicap. Monty thought about using Xen for virtualization, but he went with VMware in the end because it was more expedient, and less risky.

The third lesson is that you can outsource just about everything. Their data center is outsourced. Their HR is outsourced. Their CRM is outsourced. Four of their developers are consultants in India. Their loan docs are scanned into the system by a couple of guys who were sent over by their imaging outsourcer. The ability to outsource so much allows a startup to be up and running in a very short time, and to conserve capital by paying at the point of use. To me, the lesson for Sun is that the data center outsourcing business deserves some serious study. If we can do a Black Box so well, why not a Black Building?

The fourth lesson is that the hypervisor is the new operating system. Both VMware and Xen run on bare metal. All of the startup's applications run in separate instances of Windows, but they all run on the same HP box running VMware, not HP-UX. Sun's support of Solaris on VMware and Xen couldn't have come at a better time. IT theorists make much of the idea that cost-saving server consolidation is behind the explosion in virtualization, but my hunch is that the real reason, at least for a startup, is risk mitigation again. You don't wan't the crash of one application to bring down the whole system.

The fifth lesson is that big ticket IT gear is still sold by human beings, not bought from a catalog. Sun tried hard to get Monty's business, but HP tried just a bit harder. Again, I think that risk mitigation is at the heart of this. Monty wound up buying the hardware through an HP channel partner, and one reason was that the channel partner offered some very valuable consulting services that helped ensure the project's success.

The sixth lesson is that speed is critical. Every day without income burns another tranche of the startup's limited capital, and if you burn through too much of it, you're finished. Running out of capital is the biggest risk of all. Monty made some choices based on what could be up and running quickly, and products and vendors that couldn't deliver results quickly were cast aside.  

In the end, it's all about risk. The startup that controls risk effectively will have a better chance of survival, and the vendor that can speedily provide the products and services that mitigate risk will have the best chance at selling to this type of customer.

Sunday Apr 08, 2007

Diary of a startup - Weeks 13 and 14: hardware in the house!

The first round of hardware has arrived, a major milestone. "Monty," the CIO, has his crew working on setting them up. There are four major pieces of equipment: two Hewlett-Packard DL360 servers, each with two Xeon quad-core processors, an HP storage server, and an HP tape library that holds 24 tapes. They're rackable, with a 1U form factor on the servers, and 2U on the tape library. I've never seen a tape library up close, and I'd like to see this one. It's got a magazine on either side, two tapes high, and and a little robot that scoots up and down between them to change tapes.

Thw whole setup fits neatly on a 3/4-sized wheeled rack, which will make it easy to move to their co-located data center once it's configured. They will be running Red Hat, not HP-UX. As I noted earlier, each major application will run in its own virtual machine. The original idea was to use VMware to accomplish this, but now that the crew has had time to play around with Red Hat, they think they might just go with Xen. Score another one for open source!

I find some irony in the contrast between the orderly way in which Monty goes about setting up this infrastructure, and the chaotic state of the market that the startup will be entering. They will be starting to make sub-prime loans in the middle of the vast crisis in subprime lending, where major lenders are going bankrupt right and left. They're hoping that the massive failures of their competitors will open up a wide niche that they can exploit, but like any startup, it's a gamble. At least they can take heart that one element of risk will be removed by their carefully-considered IT setup.

Update: I mistakenly referred to the DL 360's processors as being from AMD when I first posted this entry. Thanks to a comment from alert reader zdzichu, I have corrected the post. The DL 360 uses Intel Xeon processors.

Sunday Mar 25, 2007

Diary of a startup - Weeks 11 and 12: the Age of Outsourcing

This is the Age of Outsourcing. Sun "makes" computers, yet it doesn't. There's no chip foundry, very little assembly, and very little shipping. It's all outsourced to hyper-efficient specialist firms. Even the loading of software onto the machines is outsourced.

Outsourcing makes even more sense for a startup. It avoids heavy capital expenditures, and it's quick to get up and running. "Monty," the startup CIO, regards his basic problem as integrating a bunch of purchased software with a bunch of outsourced services, then setting up a business process management system. In a situation like this, service oriented architecture (SOA) really comes into its own. The services are not just conventional software with SOA wrappers, but also actual human-provided services, some in Dallas, some local, but all off-site. From scanning loan applications to drawing up the note and deed of trust, it's all outsourced.

In this sense, the startup CIO is very much like a movie producer. He orchestrates the services of multiple independent contractors -- individuals  and corporations -- to produce a very complex product.

Asking "What does a CIO really produce?" is like asking the old Hollywood question: "What does a producer produce?" The answer is simple: results.

Sunday Mar 11, 2007

Diary of a startup - Week 10: the future is calling

It's good for those of us on the vendor side of the IT industry to hang out with guys on the user side. If you want to sell stuff to people, it helps to know how they operate. Checking in regularly with "Monty", the startup CIO, has given me a boatload of  insights into the world of our customers, and often those insights are quite unexpected.

The latest thing to hit me is that the role of the modern CIO encompasses not just computers, but also the telephone system. I guess this really shouldn't have been a surprise, since computers communicate over phone lines. "The network is the computer," right? Right. But it's more than that. The Internet and the world telephone system are rapidly becoming indistinguishable from each other. Voice is now just another form of data to be turned into ones and zeros and sent down the pipe.

It's all very well to pontificate at a grand level (see above) about the convergence of voice and data over the Internet, but it's fascinating when you can see it played out street level, right in front of you. Monty had mentioned that the co-location facility that will house the startup's machines was owned by a major telco. Why would a telco want to be in the co-located data center business, I wondered? Simple: because connection to the world network by phone line is now the most essential attribute of  the computer.  True, the co-location facility provides a secure place for the boxes to sit. But more importantly, it provides an ultra-secure connection to the network. It has multiple, redundant cable entries, at different parts of the building, so that physical damage to one cable will result in instantaneous fail-over to another one.

The startup is a wholesale mortage bank, and it's vital for its computers to stay up, and stay connected to the Internet, since mortgage submissions will flow in electronically from its stable of mortgage brokers. No Internet connection, no business. Hence the highly-redundant network connections at the co-location facility. But the submissions must be reviewed by human beings at the company's office, away from the data center, so Monty is also overseeing the provision of a dedicated phone circuit between the data center and the office, allowing the office desktops to reliably connect to the servers.

Once you've gone to these lengths to protect the flow of your data, it's logical to use the same infrastructure to protect your voice communication, by simply turning it into ones and zeroes. When a mortgage broker calls the company, the call will be answered by an automated PBX at the co-location facility. The voice will be converted by a "bridge" into Internet Protocol data, and sent over the same dedicated data line to the company office. At the office, all the phones will be Voice Over Internet Protocol phones, plugged directly into the corporate LAN. No longer does an enterprise need one set of wires for voice and another for data -- it's all a single symphony of dancing bits.

The future is arriving much faster than I once thought it would. It wasn't that long ago that I made a payphone call from the last telephone system in the United States that didn't have dials -- you just picked up the phone and told the operator what number you wanted. Any guesses where that was? I'll give you a hint: it's roughly 26 miles from Los Angeles.




« July 2016