Peter Christy

C-Scape 2008

For the last couple of years Cisco has fooled around with the format of the Industry Analyst meeting (now called C-Scape). Three years ago it was joint Financial and Industry analysts - a format I liked but apparently few others did. The Industry-Analyst-only version became less crisp (more focused on the "dialog" which isn't always a great idea). This year's meeting was transitional. Cisco hired a new head of Analyst Relations -- Melissa Selcher - from Sun, and then Melissa promptly went on maternity leave and just got back to Cisco a few weeks before C-Scape so C-Scape 2008 was driven by Terry Anderson (Mel's boss) and returned to the crispness of earlier versions, but sadly lacked any discussion of the current business (the base). There were some great customer presentations including the head of IT for the HSBC bank (the HSBC IT organization has 100,000 employees and an outside spend of $6B/year!) and the CTO of Verizon. Randy Pond, Cisco EVP of Operations, Processes and Systems gave a really clear talk on Cisco's decade long evolution to its current, unique "matrix" form of organization (Councils and Boards). And of course an exhausting set of deeper dive presentations and 1::1 briefing opportunities. Like Microsoft, it's really hard to keep up with Cisco since so much is going on in so many dimensions, but it's always fun to try!

Video and the Network (redoux)

The overarching theme at C-Scape was Video as it has been repeatedly in the past. Video of course drives huge increases in network traffic and demands for high quality network services, so for a network equipment vendor it must be an important topic. What is different this year is: (1) Cisco products are starting to drive video loads directly (TelePresence) and (2) Cisco has introduced a family of video network products. Ultimately video end-point devices will include meta-data by which they can identify their needs to the network (bandwidth, quality of service - in the interim Cisco will let switch ports act as a surrogate for the connected device). The first notable product is a transcoder that transforms common video formats - Cisco's goals is that any media can be experienced from any endpoint (a future version will do real-time transcoding). Video is integrated with WebEx already. There is no doubt that video will play an important role in the network (anyone who has experienced Cisco TelePresence or HP Halo Room will agree). The more interesting question is "how much?" Cisco hasn't tried to answer that yet. While there are some jobs (like John Chambers') that are defined by non-stop meetings, I believe that for most people trying to get a job done the goal is fewer, more-effective meetings, not more meetings. That suggests a bound to how much video a normal organization will drive.

A Bad Month for the Good Guys

For years I've been railing about the lack of real joint industry progress on Spam control and saying we're "fiddling while Rome burns." I'm starting to wonder if that was an understatement. It's been an interesting month or so for security. First a San Jose (CA) hoster was shut down because of investigative reporting by the Washington Post and the result was a dramatic drop in Spam (more than 50%) apparently because this site hosted much of the payment system used by the bad guys (the infrastructure under renting a bot net for example). The good news was the unexpected magnitude of the impact; the bad news was the fact that a newspaper was the driver. Then a couple of weeks later John Markoff wrote a pessimistic piece in the NYT suggesting that economics of malware were now so robust that the bad guys were outspending the good guys (follow the money!). And then this week Microsoft went to heroic overtime after an attack was found in the wild that was used for "drive by" infection (just visit a bad site) that was able to do anything the unfortunate browser user could do (ANYTHING). Within just a couple of days Microsoft had patched the problem. Before you start dumping on Microsoft read the Secure Development Lifecycle (SDL) blog entry Michael Howard just posted. SDL is generally accepted by experts as the state-of-the-art in trying to develop secure software; Michael's entry explains why it wasn't enough in this case. Sure feels like we're fiddling while Rome burns (we never really do much to mitigate bot attacks because that requires complex industry technical and financial cooperation). This is all very scary!

Learning to Love Microsoft

It's truly amazing to think of all the effort in the computer industry that has been driven by hatred of Microsoft, either at the company level (all the company consortia to free the world from their evil domination) or at the grass roots level (most recent CS grads know that Microsoft is both evil and incompetent). So what kind of fool would sign up to get people to like Microsoft better? The answer is Microsoft CVP Dan'l Lewin, and he may be crazy like a fox but he's no fool. Dan'l does a bunch of interesting things from his Mountain View office but a major part of his mission has always been to improve Microsoft's relations with the outside tech community, especially with the Silicon Valley infrastructure. He rolled out the latest wave in these efforts in conjunction with the Web 2.0 Conference in SF -- BizSpark.  BizSpark is Microsoft's answer to Amway in a sense, a network of partners (including IRG) that are empowered to offer relevant startups (yes there are rules) free to inexpensive access to Microsoft software for development and early market purposes. In the past, Dan'l built and still operates an organization that understands and nurtures external startups even more intimately but on a much smaller scale. I would guess that a lot of recent CS grads still think that Microsoft is evil and incompetent but from what we hear Dan'l has made real progress pivoting these companies at the management level. It's a nasty job but really valuable to Microsoft.

Google GreenTech

I had a chance to have lunch with my old friend Bill Weihl, Google's GreenTech Tzar (prior Bill was CTO at Akamai). I asked Bill what Google brought to bear in this space (compared to the likes of Chevron for example) and got a remarkably good answer. He pointed out that much of the current alternative energy work is misdirected in the sense that it does nothing to diminish the use of coal for electric power which has unfortunately climate impact unless some magic clean coal solution can be found (don't bet the rent money on that). Coal power costs roughly $4/KWH. Alternative power (e.g. Wind, solar) fetches $10/KWH (driven by governmentally mandated use by power companies) so most of today's alternative energy is optimized to make $8/KWH power and sell it profitably for $10. Google is focusing on solutions that could possibly beat coal prices (taking the long term view). Oh, and as Bill points out, minimize your expectations for companies like Chevron and BP. They advertise their alternative efforts a lot but if you read a recent 10Q they are oil companies (period).

The Data Center Market Matures

Do you remember back in the dot com boom when co-location hosting facilities (high-class warehouses with good power and cooling) were touted as a high-value business? One of the interesting and differentiated offerings at that time was Equinix. They offered neither warehouse space nor vertically integrated, locked in services. Instead their ideas were: (1) network neutral (let the user choose his network services provider separately) and (2) really high-quality facilities (when it came to things like security and uptime). We recently caught up with Equinix for the first time in a number of years. They have not only managed to survive the downturn (Exodus certainly didn't) but also have demonstrated steady growth. Their model seems pretty much spot on for the base level of cloud data center infrastructure (reliable and well-connected). It's still not easy to make money in "hosting" (Ellen Hancock was just plain wrong) but rising from those ashes Equinix's business model not only pays the rent but is even capable of providing enough free cash flow the drive expansion in these uncertain financial times.

Microsoft PDC Overload

There was an amazing amount of interesting new stuff at Microsoft's Professional Developer's Conference (PDC) in LA last week. Microsoft had cancelled the planned 2007 version because the time wasn't right (PDC is held episodically when major technology introductions are ready) but there was more than enough this time including deep dives into the cloud infrastructure (nee "Red Dog" and now formally Windows Azure), Windows 7 and of course the infamous Oslo.

Mr. Ozzie Comes to Los Angeles

Ray Ozzie really hit his stride as the executive MC at PDC (Ray led the keynotes on both of the first two days). At Mix08 earlier this year he was a little apologetic ("I'm not Bill Gates") and tentative (nothing formal to announce -- just vision) but all that was gone at PDC. Ozzie is now not only a reasonable alternative to Gates -- in many ways he's better. It was always thrilling in a celebrity sense to attend a Gates keynote, but sort of like attending a Who concert today -- it's no longer all about the music and the performance. Sometimes Bill seemed like he was just calling it in (for lots of possible good reasons). Often he explained things in the context of the history of Microsoft. Ray definitely wasn't calling it in and he didn't worry about explaining the Windows 3.1 history. The story was Windows Azure (Windows for the Cloud) and he was very articulate and passionate. Ray Ozzie has accomplished more than enough to rest on his laurels if he wanted to (Lotus Notes, Groove) but he clearly didn't sign up as Microsoft's Chief Software Architect (Gates' final position) as an honorarium for his earlier work. If Azure and Live Mesh succeed, the impact will swamp what Ray has done in the past. He's fully engaged, excited and passionate and it shows. Bravo!

Windows Azure

Windows Azure is breathtaking in vision and ambition: The idea is simple: create a version of Windows that runs Windows applications in the cloud at unbounded scale. Pioneers of successful Cloud applications and services (e.g. Facebook) all say that scalability is the critical design issue -- if your scalability model isn't right then success will literally kill you. Windows Azure has no less an ambition than enabling scalability while bringing forward the richness and tools in Windows programming. This isn't HotMail warmed over with a new face. Ozzie views the PC, the phone and the Cloud (all running Windows of course) as the three pillars of future applications, with Live Mesh keeping data synchronized among all the participants. Windows Azure builds on the success of Windows as an applications platform. The key question going forward is whether Microsoft can herd the huge and valuable Windows developer community toward Azure. PDC was a loud and strong beginning but only time will tell.

What's Oslo?

Anyone who has come close recently with an XML'ish topic has probably heard me ask "Do you know about Microsoft's Oslo?" Well Oslo is now much more out of the closet but in many ways still as mysterious. At the root of Oslo is the contrast of imperative and declarative expression (program logic and rules). Complexity is the great issue in all forms of modern application and system design. The wise men all believe that moving more and more application and system design into a declarative form is critical (rules and their interaction are much easier to understand than code written in C# or Java). Oslo is a unique and bold tool for declarative expression. Or is it? I keep asking the Microsoft marketing people to give me the decision maker elevator pitch for Oslo (or declarative expression more broadly) and they can't (not because it doesn't exist but rather because they haven't coalesced it). Microsoft's PDC is the ultimate geek fest -- the lingua franca of PDC is on the fly construction of working code in demos. PDC certainly wasn't the venue to give the CEO elevator pitch. The pieces of Oslo that were explained at PDC are interesting to a programmer but I'm sure completely cryptic to anyone on the business side. People very much "in the know" had suggested that Oslo would be used for rule based virtual machine deployment and that didn't see the light of day (and now those in the know are less confident). I guess I'll just have to wait some more and see what transpires.

Windows 7

 Steven Sinofsky gave a very upbeat introduction to Windows 7. Windows 7 of course is the version of Windows that comes after Vista, and Vista of course is one of Microsoft's biggest business embarrassments. Vista isn't at all a bad system today, what with the platform hardware catching up and with the SP1 fixes. In retrospect, Vista was just much too much stuff, all pulling in different directions. Remember that Vista was the version of Windows that (1) converged the client source tree with the Server and changed device models; (2) introduced a whole new application model (Dot Net); (3) introduced a new UI model that leveraged GPU hardware; (4) was reengineered to provide much better security; and so on. Possibly it was just too much, but certainly it was too much for Microsoft to wrangle successfully into something that was ready for the marketplace and that the marketplace was in turn ready for. When you talk to insiders, there was no shortage of contributing problems. But the causality of the Vista debacle isn't really the issue; the issue is the brand damage that occurred. Windows 7 is an entirely different beast. With the possible exception of multi-touch integration (which is very much optional), it isn't really a different or new operating system -- rather it's a much refined and better system in a lot of little ways. Julie Larson Green, who drove the UI improvements in Office 2007, has been hard at work with Windows 7 with similar results -- significantly better usability and a more pleasing user experience. And Windows 7 intends to be smaller and faster as well which will be welcome too. It took a lot of hard work by people like Sinofsky and and Jon DeVann (who we are told gets a lot of credits for herding the many engineering pieces back into a team). The product team seems to be doing a great job building on Vista and remedying the things that went wrong there. At some point soon the marketing team gets the more difficult challenge of reestablishing the brand.

Arista -- Cloud Networking

Concurrently with PDC (and we would guess coincidentally with), Andy Bechtolsheim's 10G switch company reorganized, renamed and repositioned (might as well get it all out at the same time). The big news of the reorganization was that Jayshree Ullal, most recently SVP of the Cisco data center networking efforts was named as president and CEO. It's a big win for Arista, and lets Andy focus more of his effort on the product, his passion. Arista used to be Arastra. Both names apparently have meaning for people more erudite and educated than me. Arista is thought to be simpler and easier to remember. I'll continue to think of this as "Andy B's company" (and now Jayshree's as well). I don't really care what they call it (nor do their customers I suspect). Finally, Arista positioned itself at the center of Cloud Networking, a nice story since it emphasizes the performance, space, cost and power elegance in the switch and delays the time when Arista will get sucked down into the morass of details and compatibilities that characterizes traditional networking. As we've noted before, for those who quickly recognize Andy's fingerprints, Arista is quite different since it fastidiously avoids custom parts and builds upon standard vendor parts. This time around it's all about software (in addition to the small, low-power and fast elegance of the box). Arista thinks that they can advance the state of the art in networking software as significantly as Juniper did with JunOS compared to Cisco's IOS of the time. Arista's EOS strives to be better in the important dimensions of manageability, availability (upgrade without disruption; few failures can bring down the switch) and maybe most uniquely the extensibility of the software. Andy's fellow traveler (and investor) in Arista is long term entrepreneurial partner David Cheriton, a system expert and professor at Stanford. Arista is Andy and David's Nth joint startup, and a lot of Arista core engineers are also repeat offenders. The team certainly understands software academically and understands commercial networking software in detail. If history is meaningful it's a good bet that Arista will succeed. My key question is just how flexible the extensible software is, and how far that goes optimizing important switch use cases. Time should tell.

What a Week

Paul Maritz must have had a reasonable idea of what he was getting into. In most aspects VMWorld was been a smashing success -- large and enthusiastic crowds, a big ecosystem pavilion. Paul did an excellent job on the keynote considering he had been on the job for all of 10 weeks (he was blessed by the fact that Diane Green was truly an awful public speaker so the hurdle was low) that dovetailed nicely with CTO Steve Herrod's excellent technical keynote the next day. Maritz sounded more like the leader of a $2B software company than VMware. The plans sound like business plans: (1) Paul admits they are really a platform vendor and they talk about a Data Center OS, (2) he put a stake in the ground that the Cloud was going to be a real business and (3) that they were getting serious about VDI (now branded as vClient). All of this makes sense if you're trying to answer the question "How will VMware make money?" (the question that Diane never wanted to talk about it). On the other hand, with the collapse of the "virtualization" reality distortion field came a further (and I'm sure painful) deflation of the stock, taking it down well under the initial IPO price. Some of the commentators apparently never realized that VMware really does compete with the big boys. Oh well...

Microsoft Bangs the Virtual Drum

In what is bound to be only the first round in loud and endless rolling thunder, Microsoft officially released virtualization. Nothing earthshaking was new. The standalone bare metal hypervisor was released earlier than most expected. Hyper-VMotion was demonstrated and seemed to work. Mostly I think it was Microsoft starting to show some of their powerful assets like customers, business partners and the ability to spend $100M on marketing (or whatever) when they want to. As we've been saying for some time -- Game on! VMware better forget the healthy and eco-friendly lifestyle and learn quickly to be street fighters (maybe Oracle will buy VMW -- that would make the game fun!)

Get on the Right Curve!

A treat at the Citrix analyst meeting was talk from Geir Ramleth, the Bechtel CIO and a Citrix customer. Ramleth described the fascinating saga of Bechtel's modernization of their IT infrastructure. As Bechtel was contemplating how to get from where they were to where they wanted to be, they came to the realization that big Internet players bought bandwidth on an entirely different price scale from giant Bechtel, and that Internet storage providers had a cost structure that bore no resemblance to the enterprise storage prices they paid. Geir reasoned that almost all of the VC IT investment was aimed at the consumer side (and presumably the innovation) and that if he wanted to really make progress he needed somehow to get off the enterprise IT curve and onto the consumer Internet curve. The details are pretty interesting but we're sworn to secrecy.