Visit Your Local PBS Station PBS Home PBS Home Programs A-Z TV Schedules Watch Video Donate Shop PBS Search PBS
I, Cringely - The Survival of the Nerdiest with Robert X. Cringely
Search I,Cringely:

The Pulpit
The Pulpit

<< [ Repeal Denied ]   |  End of Life  |   [ The Men Behind the Curtain ] >>

Weekly Column

End of Life: Cisco reinvents the mainframe

Status: [CLOSED] comments (44)
By Robert X. Cringely

As my son Cole, who is three years old, explains it, nothing lives forever except for vampire robots, a particular obsession of his. While I can't speak to vampire robots, when it comes to computer and networking equipment there typically is a finite life span after which vendors don't usually provide much, if any, support. It's not that the old stuff suddenly goes bad, it's that we're supposed to buy new, whether we want to -- or even need to -- or not. They call it EOL -- End of Life -- and it represents to the sales department a giddy combination of possibility and peril where, like passing Go in Monopoly, everything is suddenly new again but there is always the risk that new stuff will have on it the label of your competitor.

This week, then, Cisco Systems announced a new class of enterprise switches called the Nexus 7000 intended to replace its Catalyst router family, which is reaching its End of Life. To me the Nexus 7000, which costs from $75,000-$200,000, looks a heck of a lot like a mainframe computer. To Cisco it looks like a frigging gold mine.

These chances to tell customers they should throw out their perfectly fine equipment come along rarely, and in this case the opportunity to throw out the old and replace with new is particularly huge and gratifying because there is so much of the old stuff to get rid of. The equipment that will be replaced with Nexus 7000 racks was generally installed during the glory days of 1999-2000, when dot-coms and V.90 modems ruled the world, there was little streaming video, and we didn't really buy all that much stuff over the Internet. In anticipation of future growth back then (and just because they could), companies like Cisco pushed so much network hardware into the sales channel that it has taken until now for most of that equipment to finally become obsolete. So now they can push a boatload of new equipment into data centers in exactly the same way.

I'm not saying the Nexus 7000 is not needed or that it is bad in any way. Quite the contrary. Cisco has spent four years and $1 billion building a new generation of routers with new capabilities that are intended to be so compelling they'll keep customers from jumping to Juniper or some other competitor. And along with ensuring customer loyalty, the Nexus 7000s that start rolling out shortly will eventually enable whole new kinds of data services, most importantly robust IP multicasting as I described in this space a few weeks ago.

One huge difference between the Nexus and Catalyst lines, for example, is that Nexus comes with IP multicast turned "on," while Catalyst came with multicast turned "off" as a default. A Nexus 7000 chassis can pump up to 15 terabits per second, which is a heck of a lot of bits. Just for example, if we imagine a DVD-quality H.264 video stream running typically at one megabit per second, that Nexus 7000 could seemingly support 15 MILLION such data streams. In practical service, however, where the Nexus 7000 would be providing bandwidth for storage and network management in addition to pure file service, it is more reasonable to expect a fully tricked-out Nexus 7000 to support more like one million or so concurrent users. It is difficult at this point to even estimate the total cost of that tricked-out Nexus loaded with 10-gigabit-per-second network cards and hundreds of terabytes of storage, but it will undoubtedly set a new low cost point for per-subscriber hardware. Cisco is going to sell a lot of these puppies to telephone companies upgrading their DSL plants to offer IP TV.

What strikes me from reading the Nexus specs and that of the associated NX-OS operating system is how this new switch reminds me of an old mainframe. Nearly all services are virtualized, with multiple copies of the OS starting and stopping as needed. Everything is redundant, isolated, and intended for nonstop service. It is hard to imagine when, if ever, you'd even need to reboot. And while the Nexus supports network connections up to 10 gigabits per second, the really fast networking takes place in parallel between cards over a passive backplane. The Nexus 7000 is a data center in a rack, only with dramatically reduced cooling and power requirements which suggest to me that Cisco has a growth strategy for this architecture that will, over time, make it look more and more like a big computer and less like a router. Throw on a virtualized AIX or Solaris and the Nexus will eventually reveal that its true competition is less likely to be Juniper than it is IBM, HP, and Sun.

Remember this new platform has to last for a decade. From today's perspective making it still attractive 10 years from now requires subsuming as many computing services as one can imagine, not just undermining cable TV.

And speaking of undermining, many readers have been asking me to put in context IBM's recent move to cut pay for almost 8,000 service and support employees. I have resisted commenting to this point mainly because I see my job here as covering stories that AREN'T being handled well (or at all) elsewhere. But in the case of this story the Associated Press and others have done a good job of explaining the problem from the perspective of the employees, so I haven't had to.

But readers keep asking and there does seem to be an arm's length view of the situation that hasn't been well explained to date, so here goes.

If you aren't familiar with the story, IBM was sued several years ago by employees who were classified as exempt and therefore not entitled to overtime pay, yet those employees felt that had they worked at some other company their duties would have been considered non-exempt. IBM lost the case, paid a $65 million settlement in 2006, but took until now to decide that it ought to reduce by 15 percent the base pay of the affected employees in order to keep the settlement revenue-neutral for the company. If IBM had to pay overtime, it would tie that overtime to a lower base pay, thus keeping its costs steady.

While this probably makes total sense in the IBM accounting department, the change was a surprise to the affected workers, who say they are hurt by the lower base since it also cuts their vacation pay and IBM's contribution to their 401K. It might be easy to point to that $65 million settlement as making up for some of this, except that many IBM employees who were eligible to participate in the settlement for some reason didn't sign up for it and no longer can. Now there's a communication problem that needs exploration.

What the big picture shows here is the apparent end of IBM's tradition of respect for the individual. For most of its corporate history IBM has been a pioneer -- a model -- for corporate responsibility, but that era seems to be over. Maybe there is no more fat left to trim so the company is cutting muscle, instead. But I think there is more to it than that. I think this is a logical eventuality of IBM becoming a truly global corporation, not just an American company that does business abroad.

Despite the dark stories I have written about IBM over the last couple years, the company's latest financial reports were very good and the earnings guidance it gave to Wall Street was positively glowing. This makes little sense looking at the company from a U.S. perspective, where customers are upset and profits appear to be fleeting. Cutting through the recent IBM financials shows, in fact, that the company makes little or no money in the U.S. and quite a bit of money internationally. Nearly all of IBM's current profit, in fact, can be attributed to a single condition -- the weak dollar. International sales and profits are bigger mainly because the dollar is so much smaller than it used to be -- a condition that is likely to continue, hence the glowing earnings forecast.

Maybe what IBM is doing is turning itself into a business that is mainly NOT in the U.S. Those rosy forecasts could be based on an active plan to essentially abandon the bottom of the U.S. market in favor of the top of every international market. It hurts the U.S. employees (especially those in services) but makes sense in so many ways. The model it scarily reminds me of is Tyco, which went so far as to switch its incorporation to Bermuda.

And what if this strategy fails or the dollar recovers? Then they'll ramp up production of those vampire robots, I'm sure.

Comments from the Tribe

Status: [CLOSED] read all comments (44)

As an employee I cannot wait for ibm US to sell remaining assets and depart the US permanently, including the executives and their families. Welcome to the 3rd world. Wen you need top-notch employees in America again, good luck. I am ashamed and embarrased to be an ibm employee and am just buying my time(doing very little...catch me if you can)to collect my measley pension.

BluePigSlave | Feb 13, 2008 | 7:40PM


you certainly have spawned a lot of comment, which is a good sign :).

One area I would like to clarify is that the Cisco Nexus 7000 is not a replacement for the Catalyst 6500. The Catalyst is not End-of-Life, and, in fact, the roadmap and related investment in the 6500 still extends for several years. The logic behind the Cisco Nexus family is based on the changes we see in the data center and actually ties back to your mainframe comment.

We see the mainframe or server being fully deconstructed. Where everything was once in one box, we are pulling it apart with the advent of SANs, compute clusters, server virtualization, etc. so an application environment is drawn from pools of data center resources (compute, storage, I/O). In this environment, the data center network now needs operational characteristics similar to a server or mainframe bus. This is where the Nexus comes in--not just to deliver quantitative improvements to the data center network but also qualitative improvement to the data center network and related operations.

Omar Sultan | Feb 14, 2008 | 3:09AM

IBM is interested in globalization, not nationalization. This means having an average salary range for similar jobs and bands. The people I know that have been affected by the pay cut received little notice in advance. Is this fair? I don't think so. At the core of this are the customers. Take a poll and look for happy IBM customers, they are hard to find. IBM services used to be the gold standard and those days are gone. Essentially, they are focusing on global economies, and taking advantage of the overseas markets. This, while still under protection of being listed as a US company. Stock price is not a measure of long term health. People need to look at other indicators too.

lklllkl | Feb 14, 2008 | 6:47AM