July 23rd, 2007
Salesforce: David - When there's only 5 systems left in the world, make sure you take a PaaS
Last week, I posted a video interview, podcast, and blog about Sun’s Redshift computing theory-cum-marketing push. Redshift computing is about the sort of scale that servers will one day reach — so much so that, as Sun’s execs put it, there may only be five computers left in the world (but let’s call them systems since it’s not clear if they’ll be discrete computers as we know them). Simply put, this is the hardware take on Web 2.0.
Should both Moore’s Law and Web 2.0, including software delivered as services (everything from CRM to desktop productivity) continue on their trajectories, a future where the number of servers delivering those services is greatly reduced (compared to today) and the client (mobile, desktop, or otherwise) becomes more a renderer of those service rather than a computer is logically quite possible. The implications are pretty staggering. For example, whereas untold millions of computers are currently managing the global collective gazilloflops of compute power, a world where that’s reduced to five essentially means that compute power will essentially outstrip demand. In fact, that’s probably the case now.
If there was some way of determining the total compute capacity of all the world’s systems, we’d probably find that 5 percent or less of it is getting used on average. Meanwhile, the remaining 95 percent of underutilization is over utilizing the world’s natural resources. In other words, if we can get down to five computers, who are we to argue?
Getting there of course is fraught with all sorts of barriers. Perhaps the biggest one of those is security. I’m having a hard time imagining a major airline, bank getting comfortable with the idea that they’re just one of many other tenants (some of which are their competitors) running on the same system(s). Not to mention the sorts of organizations that believe secrecy is a key to their success (biotechs, government agencies, etc.) .
Picture for example the world’s various intelligence agencies sharing one system. OK, so maybe it won’t be five systems. But it will most definitely be fewer which means gear makers (servers, storage, etc.) will be competing for an ever shrinking market. Shrinking in terms of volume and dollars. After all, if Moore’s Law is right, not only will we be heading in the direction of five systems (or 10, 20, or whatever that final number is), but those systems in total will cost a fraction of what’s been invested in the 5 percent of compute power that gets utilized today.
As Sun VP of Redshift and Web Peter Ulander said in my video interview of him, he thinks Sun can win in this environment. But, based on a letter that I received from salesforce.com platform evangelist Peter Coffee, I’m not so sure. Coffee didn’t necessarily challenge the idea that one or more of the last five systems on the planet will be running on Sun’s gear. Instead, he noted that at least two of them would be running the salesforce.com platform and it provoked me to think hard about what sort of software services future platform providers must offer in order to best service customers, and ultimately to win.
Last October, shortly after salesforce.com started pushing its Apex platform, I pointed out what makes that platform truly unique. As far as I know, it’s the only business computing cloud that can host code developed by you. In other words, you can write software that taps the very business-oriented APIs of Apex and that code can run on salesforce.com’s systems (instead of your own). The key advantage of this approach is that scalability and reliability — the stuff associated with running hardware — are not your problem. They’re salesforce.com’s.
You could argue that this is no different than what Amazon is doing with EC2 or what the outcome would be if you turned to a Windows or Linux hosting outfit. But it is different. Whereas systems delivered from those environments have no intrinsic business value, the Apex platform is loaded with business oriented APIs for lifeblood functions like salesforce automation and customer relationship management. With salesforce.com, you don’t need to load a physical or virtual Intel box with your own business software, only to take on the headaches that follow as well.
In his e-mail, Coffee pointed out that salesforce.com is casting this architecture as a kissing cousin to Software as a Service (SaaS) called Platform as a Service (PaaS). The acronym seemed so obvious to me that I Googled it to see if anyone else had ever used it (searchstring: PaaS “platform as a service” -salesforce). Without the “-salesforce” parameter, there are about 24,500 search results. With the “-salesforce” parameter, Google reports 14 (there are only two pages) and the only other vendor to officially propose PaaS was Intel as a part of its Active Management Technology (AMT) marketing. AMT is the active ingredient in Intel’s vPro brand (extensive text, still image, audio, and video coverage from me, here). (Ironically, last week, there was some news that Intel would be de-emphasizing both the Viiv and vPro brands in an effort to simplify its outbound messaging).
What’s the relevance of PaaS, which Coffee has blogged about, to Redshift? As trends go, the idea of PaaS is perhaps the greatest challenge to Sun’s notion of Redshift. In my interview with Ulander, he talks about how it won’t just be Sun’s ability to deliver scalable gear on the hardware front, but also its ability to deliver software services. I pushed back on the interview then, and after considering salesforce.com’s notion of PaaS, I realize I could have pushed back even harder.
Sure, Sun can deliver great software on top of its hardware and it can sell and maintain that software on service-oriented terms. But longer term, the idea of PaaS raises the question of whether anybody will really be interested in the business value of Sun’s software portfolio. That’s because that portfolio is a lot like other software portfolios (eg: Microsoft’s Windows): there’s no intrinsic business value to it. A call to an application programming interface (APIs) in most of Sun’s software products does something very mechanical (and that’s somewhat commoditized these days). It draws a window on a screen, spawns a HTTP thread, checks the credentials of some user or process, or opens up a network port. Important stuff that we can’t do without? Absolutely.
But increasingly, it seems as though those very system-oriented APIs are becoming heavily commoditized and the interest now lies in APIs that, with one just a couple lines of code, result in super-serious business value. For example, mashing up territory sales data from salesforce.com with a Google map. Readily accessible APIs for that sort of business value don’t exist in the software portfolios of Sun, HP, IBM, Dell, Intel, AMD, EMC, and any other gearmaker that must figure out how to survive redshifting. Not to mention the more traditional of Microsoft’s software. They (the APIs) do exist in platforms from outfits like salesforce.com, eBay, Amazon, Google, and others. But of them, only one that I know of — salesforce.com — has unleashed its business value through a series of on-demand APIs that users can call upon via code that the service provider — again, salesforce.com — will host on its own systems.
There’s no doubt that the “systems portfolio” that Sun and other gearmakers have to offer can play some role in surviving redshift. But, banking on their existing software portfolios to earn them a spot in the final five is not a good bet. As more and more enterprises embrace Web 2.0, business functionality, when delivered in PaaS-style as salesforce.com is delivering Apex, is what can have the most impact on just about every important metric that’s important to a organization (including the bottom line). Therefore, in a truly redshifted world, traditional gear makers may be left with one of three choices: (1) If there’s five systems left, make sure they’re the muscle behind all five (one is not enough), (2) figure out ways to deliver business value instead of (or in addition to) infrastructure value (eg: become the next salesforce.com), or (3) some combination thereof.
Update: If salesforce.com’s PaaS ends up running on two of the last five systems, Sun is not off to a good start. Although Sun spokesperson Andreas Schwartz pointed out how salesforce.com announced in 2001 that it would be running its platform on Sun’s Starfire servers (Andreas was commenting on a blog about Redshift by Appistry founder/CTO Bob Lozano), salesforce.com CEO Marc Benioff told me today that those Sun servers are in the process of being phased out. Wrote Benioff via e-mail: “We prefer [Linux on] Dell.”
