Mad Dog 21/21: Watson, Come Here, I Want To See You
October 14, 2013 Hesh Wiener
Verizon is an $80 billion business growing that will grow to $85 billion, maybe more, next year. That makes it (in revenue) about 75 percent the size of IBM, 77 percent the size of Hewlett-Packard, twice the size of Google, half the size of Apple, a little bigger than Microsoft and about 25 percent larger than Amazon. So when Verizon says it wants to be a major player in computer services, its competitors have to pay attention.
So, too, must present and prospective customers, because Verizon doesn’t only want to be bigger. It wants to be better, too.
Basically, Verizon has been moving pretty fast. It is strong but scared, a combination that often means big corporate motivation. In January 2011, Verizon bought cloud computing provider Terremark for $1.4 billion, and last summer it bought Hughes Telematics, a services company that also offered real and virtualized server capacity but focused on the Internet of Things, such as smart and connected cars. Verizon spent more than $600 million on this deal. Add those up, and this is serious money even for a behemoth like VZ.
In early October, Verizon said it was going to add another leading edge technology to its cloud services portfolio, an Infrastructure-as-a-Service business based on SeaMicro servers. SeaMicro, now part of AMD, builds high density, low power servers designed around a widely admired switching fabric. These servers can use chips from Advanced Micro Devices, which shelled out $334 million in February 2012 to acquire SeaMicro. But as one might expect, the SeaMicro machines also support Intel Xeon and Atom chips.
Verizon won’t be alone. Highly regarded Rackspace will, for a sum, support SeaMicro-based private clouds running OpenStack, although it has no plans to use the machines in its own data centers. Dell is also a reseller of SeaMicro machines.
SeaMicro technology is part of a new generation of hardware that gives customers a chance to avoid or reduce the overhead and eccentricity of highly virtualized environments. HP is expected to offer its own approach, a family of servers called Moonshot. The idea here stands in stark contrast to the technical strategy provided by IBM’s big Power systems and their mainframe cousins and a distinct alternative to X86 boxes based on groups of high-end Xeon chips.
These big engine servers use virtualization to squeeze the most work out of a given amount of hardware, while the SeaMicro and Moonshot boxes are based on the idea that the best way to deliver server power is by lashing together large number of small processors. In both cases the intention is to reduce wasted power, heat, space, cooling, and hardware cost. And both these approaches as well as other alternatives will all have their advocates, because for every server architecture there are some workloads that just run great and some that beg to be hosted on a different type of platform.
The new servers and the companies like Verizon seeking to build new services businesses on them are not direct rivals of legacy providers of hardware or services. The glass house systems used for back-end services by megacorps and giant government agencies, the kinds of machines IBM is deservedly famous for and the services offerings Big Blue layers on top of them, are not at all threatened by cloudfuls of small engines. But Verizon, Amazon, Google, HP, and others couldn’t care less about bean-counting for bankers–not right now, anyway. They see something new in servers that are in some ways a mirror of the myriad tiny clients that are filling people’s hands, managing smart homes, helping pilot cars and, in Santander, Spain, a really impressive smart city, keeping an electronic eye on parking spaces.
They may or may not know what the future has in store, but they are pretty certain that the kinds of servers that will be in demand and the nature of the services companies deploying these servers as cloud systems are not going to be the types of machines they might have used a couple years ago and neither will they be the kinds of machines used now to power a Google search or an Amazon purchase or smart(ish) car highway toll-paying application. The leading services companies, the ones growing fastest, are thinking about servers that act like colonies of insects. Legacy computer systems are often organized into hierarchies or serial structure like industrial assembly lines.
By adopting SeaMicro server technology, Verizon is in a way reaching all the way back to its roots in the dawn of telephony. Alexander Graham Bell, the Scotsman who became an American, the teacher of the deaf for whom electronic invention was as much a diversion as a central theme in life, set out to create an alternative to telegraphy, a system that could transmit sound the way the telegraph transmitted text.
When Bell first got the telephone to work his first test of the device sent his voice over a wire to a receiver manned by his talented technician Thomas A. Watson. When Bell spoke, his phrase proved his concept and at the same time put the telephone in perspective. Bell’s vision of the telephone was so accurate that today, more than 125 years later, it still makes perfect sense. Bell’s words were, “Mr. Watson, come here, I want to see you.” So, in a way, Bell not only showed that his telephone worked but also that it was a waystation between the telegraph of his day and the videophone (and videophone app) of ours.
The invention of the telephone was a lot more complicated than a tale of a brilliant tinker and his vision. It involved a race among inventors to patent offices, which Bell won, and subsequent litigation that occupied a lot of Bell’s time (and that of his lawyers) for many years. Bell prevailed, but in the end was happy getting out of electronics and back to his earlier life as a teacher of the deaf.
Watson didn’t stick with the telephone business either. After he left Bell’s company, he ended up becoming wildly successful in shipbuilding. He is also known for bringing the Indian mystic and philosopher, Meher Baba to the United States. Baba is perhaps best known for his catchphrase, “Don’t worry, be happy.”
Verizon, having watched Apple invent the iPhone, a product that has dominated and reshaped mobile telephony, realizes it must be part of invention if it wants to lead in any of its business segments. While Verizon may enjoy the leading mobile telephony market share in the USA, it doesn’t have the kind of power it covets. But it might be able to achieve a truly dominant position in a new market, one where, as in the long gone days of telephone before the iPhone changed the world, it could deliver something the market craves to a segment that is new and, for now at least, unbounded. That segment, it turns out, is cloud-based Infrastructure-as-a-Service, IaaS, at least according to IDC.
Recently, IDC said that the cloud services market is set to more than double during the next three years, sailing past $100 billion in annual intake. More than a quarter of that market, and arguably the fastest growing part of cloud services, will be IaaS. The two other segments in IDC’s analysis are Platform-as-a-Service, PaaS which includes the kind of hosting IBM knows very well, hosting that includes Power and mainframe servers as well as X86 iron. But that segment accounts for no more than 15 percent of the cloud market. Within PaaS the growth opportunities seem to arise not so much from legacy platforms (which may be the most profitable servers in this specialized field) but from X86 hardware configured in one of a number of popular forms to create a stack of hardware and software and virtualization.
The X86 PaaS market and the larger and closely related Software-as-a-Service, SaaS, segment, are brutally competitive. IBM discovered it could not compete in this vicious segment using its traditional technologies and marketing methods. This year it bought into what is most likely a more promising approach to the cloud by acquiring SoftLayer. IBM is very serious about this, investing $2 billion for starters, and apparently is open to learning a new way to do cloud business if that is what it takes.
But even if IBM is investing more than Verizon in the cloud, IBM has yet to move to the leading edge the way Verizon has done and the way HP will, too, when it harnesses its Moonshot servers in a way that defines an ultra-granular IaaS cloud offering. So far, IBM has announced nothing that would allow it to compete with Verizon or Rackspace in cases where the workload runs best on many small processors, jobs where virtualization gets in the way of quick and lightweight provisioning. In other words, IBM is still not ready with an offering that is the conceptual opposite of the kind of hosting at which it is so well practiced, hosting in which elaborate virtualization schemes allow a system to use nearly all its resources nearly all the time.
At issue, though, is not how to deliver services for legacy workloads. IBM is a master of that sort of work. Nor is the hot new cloud market going to be one supporting the contemporary workloads that are so well understood by Google, with its searches and advertising management, and Amazon, with its infinite virtual stores, among others. The unfettered opportunities that IBM must seek in addition to those it already understands and pursues are those that will arise as enterprising users discover the advantages of new hardware and the new offerings from the likes of Verizon, Rackspace and, I expect, HP.
One application IBM might want to host on a server that has extraordinary granularity and nearly liquid provisioning is its own Watson. It might turn out that the same kind of server architecture Verizon believes will help it address the IaaS market and which HP makes with its Moonshot line, could make Watson, with its alluring technology, affordable. All IBM needs is a granular mesh server of its own.