By davidleetodd on May 27, 2008
Check this out. I love the longhorns on top of the Sun box.
Check this out. I love the longhorns on top of the Sun box.
People often ask what the difference is between OpenOffice.org and StarOffice. This question has been popping up more often after Google's announcement that it would distribute StarOffice for free as part of the Google Pack of desktop applications.
Here's my take on this, gathered from my colleagues on the OpenOffice and StarOffice teams, especially Louis Suarez-Potts, OpenOffice.org Community Manager.
OpenOffice.org is a community Open Source project, released under the Lesser General Public License. Anyone can download either the binary or the source and redistribute them, at no charge, subject to the restrictions of the license. Sun is the biggest contributor to this project, but by no means the only one. Sun's contribution to this project is its gift to the world.
Like the Linux community, the OpenOffice.org community encourages others to develop their own distributions -- “distros” -- of the OpenOffice.org software by picking and choosing from its modules, and by adding their own modules. StarOffice is such a distro, created by taking the OpenOffice.org core, and adding various commercially-oriented modules, such as an advanced spellchecker licensed from a third party. The fact that Sun Microsystems created the StarOffice distro, and charges for it, does not mean that Sun considers StarOffice to be more “official” that OpenOffice.org, any more than Debian or Ubuntu is more “official” than the Linux kernel. StarOffice is just oriented toward a particular market segment.
There are other distros that target other markets. Sun created another distro called StarSuite that is oriented toward the Asian market. NeoOffice is an independent distro that has been ported to the Macintosh Aqua user interface. Novell created its own distro as well. The Google Pack StarOffice distro is a non-commercial version that is limited to Windows platforms.
The OpenOffice.org ecosystem encompasses all these distros, and also includes many indpendent vendors who provide support, training, consulting and other services for the various distros. Sun sells support and other services for OpenOffice.org, StarOffice and StarSuite, and cooperates with independent support vendors in various markets. We take pride in our major contribution to this ever-growing, open ecosystem.
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.
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.
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.
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.
I met "Monty," the CIO, for dinner this week. I hadn't seen him for a while, and I was struck by a certain look of fatigue that he doesn't usually have. Startups are hard work. He came alive as we talked about the company, though, and took me through the latest developments.
His hardware plans have coalesced into actual orders: two HP quad-core rack-mounted servers, one backing up the other. Running VMWare, each will host between three and five virtual machines, one for each major application. As Jonathan is fond of noting, this style of server consolidation is the wave of the future, to save both space and, more importantly, power.
Ironically, the servers will be housed at a co-location facility that Sun recommended. It's out near LAX, run by a major telco. These co-location facilities use the same strategy that the Dodgers use when they sell you a beer for eight bucks. Monty tells me that they charge $600 a month for one 20-amp circuit! Now I understand why saving power is so critical.
On the integration software side, Monty decided against Oracle Fusion, and will go with BEA middleware. Poor BEA. Little did they know that Monty was raised in a family of militant trade unionists, and absorbed negotiating tactics along with his morning cornflakes. BEA first quoted a rip-off price, but Monty countered by threatening to use an open-source stack based on JBoss, and BEA folded like a poker player with a busted flush. That's the great thing about open source -- you don't even have to use it to reap the benefits.
Week 7: I talked to Monty late in the week, and the news wasn't good for Sun. Monty is pretty sure that he'll go with HP hardware.
The reason why is somewhat nebulous. Monty sounded a little embarrassed when he told me, so he may have been trying to spare my feelings. No worries, amigo. As Michael Corleone said in The Godfather, on his way to commit a double murder, "It isn't personal. It's just business." Monty said the Sun people he dealt with were great, very knowledgeable, but -- and this is the nebulous part -- they didn't go after the business as hard as HP did. He felt that Sun was less interested in a startup than in going after big, established accounts.
Hmm. I guess this shows that trying to change the direction of a big company is like trying to steer a supertanker. You can put the wheel hard over, but it still is going to take the ship a hell of a long time to respond. Sun has been trying to point the ship at startups for some time now, but some days it's hard to see the ship actually turning. I think it also shows that Sun is a bit naive to think that startups are going to be attracted by our Web-orders-only Startup Essentials program. A startup is by definition a risky enterprise, and the startup's managers are going to look for allies to mitigate their risk. A hardware vendor can mitigate the startup's risk by providing personal advice and counsel, and hands-on assistance. We tend to think that you attract startups by offering special low prices if they order online, but the evidence suggests that risk mitigation trumps price hands down.
The software part of the deal shows the same thing. Monty is aware of Sun's Java Enterprise System and Java CAPS, and its very cheap subscription pricing attracts him, but in the end he is planning to purchase either Oracle Fusion or the BEA stack, both of which will cost him more than Sun's offering. Why? Less risk. The developers he has on staff know these products, and they don't know Sun's, so they are a lot more likely to bring home a successful project. Risk mitigation wins again.
I'm in the services end of the business, so I am biased, but I think the lesson here is that we should think less about offering discounts to startups, and more about creating services that will get them up and running quickly on our products, with a minimum of risk.
Weeks 5 and 6: It amazes me sometimes how much Monty's thinking mirrors Sun's reading of where the industry is going. Or perhaps it's more accurate to say that Sun is showing that it actually understands what's going on down at the street level where startups are put together. Most gratifying.
Monty has been thinking hard about server virtualization. His original architecture envisioned several servers, one for each major application, but this was really just a logical conception. Now that the time is coming to give the architecture a physical embodiment in silicon, those "servers" are beginning to take shape in his mind as virtual entities on one big box. When I talk to him this week, he seems to be channeling Jonathan as he describes the need to save on both space and power consumption when they get set up in a co-location facility. Or maybe Jonathan has been channeling Monty. When I close my eyes, it's hard to tell them apart.
The big question on Monty's mind now, if he decides to go with Sun hardware, is whether to go with x64 or SPARC architecture. Some of the mortgage applications he will run must be Windows-based, and he's not sure how well they would work running on virtual OSs under SPARC. He thinks it might be better to go with an architecture where the native OS is x64-based. I check out some Sun literature on the subject, and it sounds as though he's right about this one, but he really needs a Sun architect to advise him.
As a services guy, I can really see in this situation how a vendor's strong services arm can be critical in a sales situation like this. Our business is not just about shipping boxes, it's also about providing expert, honest, trustworthy advice.
Week 4: "Sally," the Sun account executive, has been busy. She invited the leadership team from the startup to one of Sun's offices in Southern California, and brought in a Sun architect to discuss their needs. In addition, Sally is working on setting them up with a possible co-location center run by a major telco, as well as with other potential partners. When I talk to "Marian," the CEO, she calls Sally one of the sharpest individuals she's met during the whole startup process. She notes that it's painfully evident that some vendors have very little interest in helping a startup that isn't going to be a $25 million customer, not responding to an inquiry for weeks.
Score one for Sun! Let's hope the momentum continues. It does make you wonder. When I step onto a car dealer's lot looking to buy a $15,000 Nissan Sentra, a sales guy is all over me like flies on barbecue. If I don't buy that day, he'll call me for weeks. Yet for a $100,000-plus server deal, a lot of vendors won't return a phone call. Thank God Sun knows better.
Week 3: Things are moving fast now. When I talk to Monty today, he's dog-tired. He got back from Dallas late last night, and is still feeling the after-effects of the trip. He flew there with Marian, the CEO, and "Dan", their project manager. They went to Dallas to check out "CIS" (not its real name, but close) one of those amazing new-breed tech companies that's bigger than Sun, yet that no one has ever heard of. At least I hadn't. More than 50,000 employees, and they're completely off the radar.
CIS does a lot of things, but Marian, Monty and Dan went to see them to talk about outsourcing the mundane details of their new mortage startup: taking the staples out of loan documents, scanning them, and converting them to indexed images. According to Monty, "All the big mortgage players use them." Who knew?
I guess I am pretty naive about how the hardware world really works. I had assumed that when Monty gets done buying the servers he needs, he will put them in his own little data center next to the broom closet. Hah! If he can come to terms with the boys in Dallas, CIS will put them on a rack in their giant data center, behind armored walls and sucking on uninterruptible power, along with the racks that belong to a thousand other companies "co-located" in the same facility. Monty may never even see the servers, because he will have them shipped there directly from Sun.
At least I hope they come from Sun. This is beginning to look a little dicey. CIS is happy to rack any kind of box, but they have a distinct preference for IBM, and don't have the Solaris expertise to be much help in adminstering the servers. So if Monty goes with Sun, he will be administering the servers himself at the end of a very long phone line, albeit a very reliable, double-terminated, hardened phone line. CIS will see to that. Also, if he lets CIS procure IBM gear, he will have the advantage of their considerable negotiating power with Big Blue.
Ironically, when I do a Google search on "CIS" and "Sun Microsystems" up comes a "Customer Success Story" about how Sun has sold a boatload of servers to CIS for its co-located hosting activities. Hmm. I guess the boys in Dallas didn't get the press release.
In addition to operating a big-as-Texas data center, the good ol' boys from CIS are no slouches when it comes to entertaining, either. They treat Monty and his colleagues to a memorable dinner at Al Biernat's legendary steakhouse. Monty is pretty worldly when it comes to fine dining, but he rates the 21-ounce, bone-in New York sirloin as the best he's ever eaten.
"Sally", the Sun account executive, is supposed to call Monty in a couple of days to take the sales process to the next level. I hope she knows how to sell against IBM.
Week 2: I talk to Monty this morning. He says the call with Sally, the Sun account executive, went well. It was mainly a get-acquainted call, and they will do a followup call this week. Marian, the CEO, was able to join for a few minutes at the end.
Monty has his work cut out for him. The problems he faces in getting the enterprise IT operation set up are non-trivial. He envisions a portal for mortgage brokers to submit their deals, but also the capability to accept paper applications, which will have to be scanned into the system. There will have to be a pricing engine, an underwriting engine, and a secondary marketing system to submit the packages to the agencies and to Wall Street. These will all come from one vendor, but other vendors will supply the middle office system and the document preparation system. The portal and the scanning system will be sourced from yet other vendors. Bottom line: a massive integration problem.
In Monty's view, the integration system is the most critical, since it ties everything together. As a former integration product manager, I think this criticality is not well appreciated in the software world. Industry pundits tend to write off integration as just another commodity. Like hell.
Monty is leaning toward Oracle Fusion middleware. Naturally, I pitch Sun's Java CAPS to him, but his mind seems to be made up. Fortunately, he is also leaning toward Solaris for the mission-critical tasks of running the database and the app server, although this is by no means settled. Sun will definitely have to earn his business. He is already thinking of Dell for the portions of the system that will have to be Windows-based. It dawns on me that he doesn't realize that Sun also offers Windows-based servers, and when I give him the news, he seems interested in the idea of having a single hardware vendor. I think to myself that if this savvy CIO doesn't know about our x64 offering, Sun's marketing efforts have a long way to go. In a sense, it appears that our constant emphasis on Solaris has put us in a classic PR trap.
I press Monty about his impression of Sun's open-source strategy. With Solaris freely available, does he intend to purchase a support contract? Yes, he does. He notes wryly, "It's hard to run an enterprise if you can't talk to the vendor." Score one for Jonathan's vision! As Monty points out, it's all about reducing risk. He wouldn't want to put the entire enterprise at risk just to save a few grand a year in support costs.
I also ask him why he doesn't go with an open source stack like Ajax. His answer surprises me. I had thought that maybe a commercial enterprise wouldn't go that far into the open source world, but he has no problem with it, and has worked with open source applications in mission-critical tasks at another organization. Rather, he sees Ajax and similar open source stacks as more suited to from-scratch development than for solving an integration problem. For integration, you still have to go commercial.
I have a hunch that integrating a bunch of relatively unknown mortage origination applications is going to be very, very tough. Any fool can integrate PeopleSoft and Siebel, but tying together vertical-specific stuff is a task that separates the walkers from the talkers. We'll see.
A long time buddy of mine, Monty (not his real name, but close), left his position as a high-level architect with a big money manager at the end of the year to join a startup as CIO. The startup, which I will call Nimble Financial, is planned as a wholesale mortgage bank. The CEO, Marian (not her real name), is a well-regarded figure in the financial services industry with strong contacts on Wall Street. The company has obtained significant funding from the Street, and intends to be funding loans by the middle of the year.
They have started operations in a warehouse in Southern California, and are looking for a deal on ten or so servers and accompanying software. With their permission, I will be documenting their progress through this blog, focusing on their IT efforts. All names will be changed to protect both the innocent and the guilty.
Week 1: Hoping to help Monty out, and garner some business for Sun, I decide to see if I can help them score some hardware. I am somewhat stunned to realize that, even though I am a Sun employee, I have no idea how to refer hardware business to Sun. I check the Sun employee web site, and can find no page devoted to referrals. I know a lot of software sales guys through my work, but no hardware people. I decide to call another friend, a popular blogger who is a Sun sales engineer, for advice. I have seen in his blog, and Jonathan's, reference to the new Startup Essentials program, but I feel that the startup may require more Sun advice than that program provides. He refers me to the Sun account executive for the area, "Sally". I call her around closing time, and she is quite friendly and helpful. She warns me that the account may be too small for Sun to handle directly, and that it may wind up being passed to a partner, but promises to contact Monty and see what Sun can do to help him.
True to her word, she emails Monty that night. They exchange emails, and set up a conference call with the two of them and Marian for the end of this week.