Facebook's New Data Center Is Bad News for Cisco

Back in high school a bunch of us ran an ISP. One day it came to pass that our main Cisco router and the access switch which housed all the dial-up modems (made by a company called Shiva iirc) stopped playing nicely together, resulting in all of our dial-up users (i.e. almost every paying customer we had) not being able to log in. Needless to say, this was a problem. Fortunately for us our ISP was technically owned by the very large metropolitan school district of which we were a part, which allowed us to use their platinum-clad service contract with Cisco tech support in order to diagnose the problem. We got the Cisco guy and the Shiva guy on the phone all at once. To cut a long story short, a) the problem was with the Shiva box, b) the Cisco guy knew more about the inner workings of that box than did their (Shiva's) own tech support, and c) the Cisco guy fixed it over the phone in 5 minutes. As an impressionable 17-year old, this display of competence left quite an impression on me :)

It helps that they have an entire educational line of courses to build up the specific skills. I took the CCNA classes (and tutored for them at a local community college) in high school and while they do go over basic topics like 7 layer model, its highly Cisco hardware specific and it helps to have funnels like that into your support operations. It is similar to how Microsoft was (still is?) in the operating systems industry.

Asking as someone with extensive familiarity with Cisco TAC, Advanced Services, et al... have you interacted with them lately? Anecdotally based on my interactions, the average knowledge level of any given set of 1st/2nd tier support engineers seems to have dropped precipitously in the past two years. Once you get to a platform-specific support group (say, the ACE group in Framingham, MA), the knowledge level is outstanding - but it also seems to have gotten much more difficult to escalate cases to that level without an extraordinary amount of effort on the part of customer engineers.

> Even if you can download it and its free and open source, most companies want someone they can call when their IT software doesn't work. Cisco isn't just a technology company. They're a giant sales and services organization.

This was also true of IBM back in the day, though.

That could be said of Microsoft in the server business as well, but Linux still managed to gain a big share. It's possible for a RedHat model to develop in networking - sell services and support, but not hardware. See Cumulus for example. (Disclaimer: I work for a Cisco competitor, but not Cumulus).

I had the fortune to speak directly with people who have been through the Cisco training program, and it is pretty rigorous. Pretty impressive how they can not just answer my question about the equipment, but put it in context with how networking "works" overall.

Regarding the Facebook move - I think it's funny how people are going to try and copy Facebook here. Facebook has a very unusual business problem they need to solve, and they are very creative and clever about solving it.

A lot of folks are going to look at Facebook from the outside, ignore the unique business problem they are solving, and say "we should insource all our network gear etc." Next thing you know they have a whole mess of new problems on their hands, support and staffing at the core. I saw this with an insourced VOIP phone system. Yes, it is cheaper but when it stops working you wind up paying for support, plus the features fall behind without care and feeding over time.

The downside for the companies that aren't Facebook (or Google, Amazon) is that you can miss out on a whole range of technology that will solve your problems just as easily, with support and the "ecosystem" to go along with it. If you don't have a unique problem (and this is part of the vanity in these decisions IMHO) then you are well served to leverage the options available.

Database maker, ha. Oracle is a much bigger sales and service organization that swallowed up Sun.

Cisco's future maybe bleak but not because we'll all move to Facebook's and Google's model. The systems they build are at massive scale, and heavily optimized to their workflow. Cisco equipment is in general designed to handle a heterogeneous workload, in a variety of deployment scenarios.

It was posted a few days ago:

https://news.ycombinator.com/item?id=8609391

I'm pretty sure he didn't even RTFA.

And, related to what he wrote, yes, you can buy a more powerful Mikrotik router for less than one with the Cisco name on it. The software on 'em sucks, though, and they have some weirds things about them. Additionally, they used to flagrantly skirt around the GPL although I'm not sure if they still do.

(not the same one I linked to, that one has a max throughput of 16Gbps while the one i linked to is max of 28Gbps) The one you linked to is MSRP $995 USD.

Still an incredible deal for what you get.

The focus on the "widget engineering" is misleading. Are these the right widgets? dunno. The bigger questions are about the widget business. The stock is under-performing the market, despite major influexs of market inflation (ie, price bias), and huge shifts in the secular demand for bit-throughput.

The company has had the leading position in the widget business for a generation. But this raises questions about the future viability of the widget business if ROI is essentially zero (?).

Is the widget business inherently not-profitable? Or is the company mismanaged (outside of widget engineering?) I don't know the answers, but it seems people should be thinking about them.

The engineering teams are mostly like not the main issue. The lack of more lucrative product/market fit seems to be a perrenial issue for the higher level exevutives.