empowering the technology luminary to re-define 'simple' and re-invent 'complete.' - Part 2
"When we set out to create a community of technical scholars in Silicon Valley, there wasn't much here and the rest of the world looked awfully big. Now a lot of the rest of the world is here"
Frederick Terman
In 1984 the Stanford Cardinal football team hired Jack Elway as their new head coach. He was the father of the already infamous Cardinal Quarterback John Elway who had been drafted by the Denver Broncos of the National Football League in 1983. After refusing to play for the Baltimore Colts, who owned the first pick of the draft and had stated their intention to draft him, Elway threatened to play baseball instead for the New York Yankees, but he was ultimately traded to Denver and the rest, as they say, is history.
stanfordquarterbacks.com
Quarterback Fred Buckley, 1984
The 1984 team finished with a dismal overall record of 5 wins against 6 losses but across the campus, not far from the football stadium, an air of excitement filled the hallways and protruded through the breezeways of the Computer Science Department building. This was not only due to the fact that Elway, in just his second pro season, had led the Broncos to their best ever NFL season record of 13 wins and 3 losses, but also because of two department operations staffers, Leonard Bosack and Sandy Lerner, who had adapted a multiple-protocol router software from a shelved piece of code written by a Stanford medical research engineer some years earlier and who were now about to embark on the launch of their company they nostalgically called Cisco Systems, intent on inventing the way networked devices would connect to and communicate with one another and whose logo was inspired by the spans of the Golden Gate Bridge
wikipedia
Cisco Co-Founder Len Bosack early business card with first logo
For a time, campus and then industry rumor had it that the two designed the first router so that they could connect incompatible computer systems in the Stanford offices at which they worked in order to be able to send romantic love letters to one another. While that’s long since been proven to have been manufactured lore, the fact does exist that on the same campus, in a similar shaped building not far from the one where Cisco’s code had it’s origin, was once the location of an office occupied by 'the father of Silicon Valley, ' Fred Terman. He was an American academician and professor at Stanford who became the dean of the School of Engineering.
He served as Provost at Stanford from 1955 to 1965. During his tenure, Terman greatly expanded the science, statistics, and engineering departments in order to win more research grants from the Department of Defense. These grants, in addition to the funds that the patented research generated, helped to catapult Stanford into the ranks of the world's first class educational institutions, as well as spurring the growth of Silicon Valley. Terman's efforts to create a mutual relationship between Stanford and the tech companies in the surrounding area also significantly contributed to this growth.
Stanford University website
Stanford University Computer Science Department
For a quarter of a century, Cisco primarily focused on one thing and they did it exceptionally well – they connected computers and their users to the internet. They were able to do this by perfecting their ‘routing’ software and hardware in an industry that connected networks of computers together, an essential requirement for fast growing businesses in the 90’s, a time when the internet and personal computing were rapidly evolving, and forever changed the way we communicated both at work and at home. They did this so well that the very name Cisco became ubiquitous for connecting anything, at home or at work to the internet; to this day, as a $50 billion American multinational technology conglomerate, they are as much as ever, synonymous with ‘the network.’
In the late nineteen-eighties, I was a field engineer on-site at Caltech in Pasadena, California responsible for the maintenance of mainframe systems under a service contract to my company that saturated the science and engineering research laden university. A substantial portion of my job involved the installation of new infrastructure, primarily miles of coaxial cable in ceilings and under raised computer room floors. We didn’t call them data centers in those days so I won’t change that description for the sake of editorial consistency.
All this cable ultimately ran to a closet where a router or a switch blinked teasingly in the dark, eagerly awaiting it’s connection to one of it’s ports, joining floors, departments, buildings and various networked devices from multiple campus locations together in what is called a LAN (local area network.)
Little did I fully appreciate at the time that this work was at the very root of a complex global web of cables, routers, and switches that would ultimately form the internet as we know it today!
Cisco may not have been the first tech company to develop and sell dedicated network devices, but it was one of the first tech companies to sell commercially successful routers that supported multiple network protocols (Chiefly a network protocol defines rules and conventions for communication between network devices. It also includes mechanisms for devices to identify and make connections with each other.)
So again, as with their recent entry into the Hyper-converged space, Cisco was not the first to market with network technology – but Cisco was successful because they were better than everyone else.
Chairman and CEO of Cisco, John Chambers, once made a few public observations on why he thought the company had continued to thrive in a very crowded market environment, noting:
“Every year there’s a new competitor or a new technology that’s going to completely leave Cisco behind. We focus on competing [against] market transitions. You get the market transitions right, and understand what that means to a customer, that’s how you win.”
Forbes
Cisco CEO John Chambers
“We’ve left behind almost every good competitor, and that’s from doing basics in ways that others have not.”
Let’s fast forward 20 years, $130 billion in market capitalization, 170 acquisitions, 9 releases of Windows Server, 2 CEO’s and 1 giant press release I remember well.
It was March of 2009 and I had just been named the North America Marketing Manager for High-end Intel Servers at IBM. IBM share in that space had been slipping to HP but we were confident of our ability to recover from that as we ramped towards launch later that year of our next generation offering outfitted with Intel’s new Nehalem microarchitecture.
Like Cisco, there were things that IBM did very well. They were for a time, never willing to compromise on design to save a buck, offering blade servers, amply outfitted with Cisco Nexus switches. Cisco was a key technology partner of IBM, and with virtualization finally getting over the adoption hump and becoming more and more of a staple in data center modernization efforts, their switches in our servers were a winning combination that allowed for the consolidation of ill-fitting hardware and software – a precursor to converged infrastructure (CI). By ill-fitting I mean systems that were serving a niche, near antiquated purpose but were still necessary to a company’s operation.
A good example of this might be that of a server occupying valuable real estate space in a rack; a server that of course has to be powered and cooled, requires a maintenance support contract, and likely has associated operational expenditures in the form of software subscription – needed for bug fixes at a minimal – and needs to be kept running in place because of a possible scenario like the following: It’s a server that’s not running mainstream Windows but rather is Linux-based and has but one key application it's hosting, an app that generates a daily 'sales out' report that is valued by the office of the CEO. There’s one big problem facing IT, the developer that wrote the app has left the company and there is currently no one in place that knows the first thing about how to replace it with something mainstream and further IT has no easy way of convincing the CEO to give up that app’s slick reporting format – so it sits – and it runs one report at the end of each working day that takes ten minutes of computing time. The other ten thousand and seventy minutes of the week the server sits idle.
This was the early promise and value of virtualization, led by a fast growing software company aptly called VMware, to eliminate that server by creating a virtual equivalent of it and host it’s operating system and niche app in a virtual machine (VM), alongside many other VM’s, also originating from similar niche servers and scenarios that had been spread across the data center over time, all now tucked nicely onto 1 well equipped virtual host server; not sitting idly – rather running at near maximum compute capacity 24/7/365.
Virtualization offers a wondrous solution to under-utilized and siloed system sprawl in a data center, but it also places extensive demands on the network. A fast, reliable switching infrastructure that is designed to provide high performance in a virtualized environment is critical to meeting next-generation data center requirements and thus was born the Cisco Nexus family of unified fabric switches that IBM recognized early on as a preferred piece of their virtual infrastructure solution incorporating them as valuable options in their BladeCenter chassis.
The networking switch fabric is the basic topology of how a network is laid out and connected to switch traffic on a data or circuit-switched network.
The concept of switch fabric, in which many network connections are made in a matrix, has a long history in communications, dating back to the circuit-switched voice telephone days when connections were switched by humans operating a switchboard.
The term 'fabric' being derived from a view of this pattern multiplied several times over by a far more complex layout of an actual switched network.
Cisco Nexus switches can support the extensive bandwidth requirements that result from aggregating servers when you consolidate or virtualize your data center network. And by helping to enable scalable server virtualization, the switches also help reduce power and cooling costs and support highly available data center core and server access.
As server consolidation projects aimed their sights at the blight of under-utilized and siloed systems in a data center, vendors like IBM who had established key partnerships with the likes of Cisco and VMware were a formidable alternative to the status quo and all seemed well in the world of customer choice for undertaking this endeavor until the unthinkable happened.
On that day in March of 2009 I was at IBM's corporate campus in Raleigh, North Carolina’s Research Triangle Park, where their server business resided, and my phone rang.
And then an urgent email arrived from my manager, followed by more calls and emails, and then I heard loud chatter in nearby cubicles, the calls, emails, and conversations all centering around the same news – Cisco had a surprise announcement for what was called their Unified Computing System, UCS for short. The fallout from their stealth entrance into the lucrative x86 server market was on us like a spaghetti stain on a white dress shirt.
‘How the heck can they do that?’ we all thought and asked each other aloud, hoping the news was inaccurate. ‘It’s crazy of Cisco to do this – it’s a path of doom and destruction their headed down – what do they know about servers - don’t they know that the field of server vendors is already crowded - don't they realize virtualization is paving the way to commoditized hardware that’s getting tougher to sell on value and differentiation - What about our partnership - They’re nuts – they should keep doing what they do best and stick to networking!’
UCS was, and still is, a simple idea that Cisco had; to further reduce their customers total cost of ownership in the face of virtualizing the data center, improving their ability to easily scale their compute footprint, to make it easier to manage a virtualized system and increase performance by integrating the different data center components – computing hardware, virtualization, switching fabric, and management software - into a platform where all can be managed as a single unit from a single pane of glass. Their system would prove to be a key component of the debut of converged infrastructure (CI). CI would ultimately become the most pedestrian term in IT since ‘plug-and-play’ [I almost wrote banal but that would be a little harsh given it’s ancestral relationship to the subject of this piece] but we’re here exploring the emergence of it’s follow-on HCI.
Experts Decision
Cisco UCS C-Series Rack Server
So Cisco found itself once again not being first to market with it’s x86 server offering, yet right out of the gate, customer adoption was prolific, as they, like many of us at IBM, recognized their leadership capability in fabric computing and CI as fundamentally sound given it's centered around their excellence in networking technology and so off they zoomed from last to third in less than two years in worldwide x86 blade server factory revenue.
To conclude part 2 here's a thought I had when researching the financial growth of Cisco over the years; if I had the foresight back in 1990, when the company held its initial public offering, to invest a grand, just a thousand bucks in their stock, I would have received 55 shares, that after adjusted for stock splits, today would represent a whopping 16,000 shares, an investment that would now be worth just under a half million dollars. Of course, I could have registered the domain name business.com and sold it in 1999 for $7.5M and retired nearly 20 years ago but that is the topic of another post I am working on.
In Part 3 we'll cover UCS groundswell and lead-up to HCI and how once again Cisco may have been last but are doing the basics in ways that others are not - and that is called Complete Hyperconvergence.