Amazon's already done this and replaced expensive Cisco Nexus routers with their home grown in house creations. Big hardware networking equipment vendors are going to continue to see this type of encroachment from large IT companies IMO.
If I recall correctly Google built its own network gear long time ago. Completely makes sense. Looking at Cisco product with regard to security & the recent NSA leaks you really don't want your network gear from them.
The big difference here is that states can simply buy access to, e.g., a Cisco Nexus and attack it from inside and out until they find a vulnerability in NX-OS, let's say, a malformed CLI-via-HTTP call.
Whereas, what software does a Google switch even run? What's the architecture, the APIs? You basically need someone inside Google, or for one of these things to fall off a truck. Way more involved and expensive than the 10k you might spend on a Nexus to throw it your lab and set your hackers on it.
Actually, Google has published papers and have presented talks (many of which are available on Youtube) on the type of gear they have developed. I don't know what their latest versions are, but recently they were using OpenFlow style infrastructure to provided fine-grained control (security, balancing, analysis) over flows through out their network. OpenFlow style constructs also provide a micro-segmentation style control (ie distributed firewall) over ingress/egress of traffic at the individual container/vm port level.
Ah shareholder value. Value is the keyword. It's not shareholder money per se. If a companies stock jumps because people think you are bold for making your own secure network equipment then you have created value. Security and privacy can make value. Look at Apple. Public perception of Apple makes up a nice part of their stock and it's based around (just to name some smaller ones) privacy and security.
Well mostly because they don't rely on any legacy code base but they can write it all by themselves. Also Google has somewhat of an reputation for security other than Cisco.
Think of it in the same terms as any other service you're contemplating: do you want to pay other people for proprietary stuff, or do you want to pay your own people to develop skills around open stuff?
cisco has an incentive to not look foolish, but they don't particularly care about any small or mid-size accounts. Your people have an incentive to care about your security and functionality. Weigh up the advantages and disadvantages and make your choice appropriately.
That's the sort of thing I'm talking about. Good implementations of known-good crypto.
I don't doubt that Google's trying to come up with new, novel cryptographic methods as well, but those take a long time to develop, test, and get reviewed. RSA didn't show up over-night and SHA3 has taken some time to get ready for production.
But Amazon doesn't want a pretty capable switch with a lot of media options; they probably want one capability (IP routing) and one or two types of media.
This Exactly. Companies like Amazon, Google, FB, Microsoft need 10% of those features. However oftentimes there's a bug in the 90% of the software that they don't use which impacts something in the 10% they do use. This causes a long delay in the roll out of fixes and creates a lot of disruption in scaling out the network. Furthermore Cisco must test all those features before they can ship a new product. This results in delay of new physical layer technology into the network. What the big 4 want is that new tech to get to market faster so they can keep up with the demand. The problem ultimately is the pace of network technology is not keeping up with the demand of the webscales.
Does this not also somewhat mean this is what Nortel should have done? What did RIM bring to the table that Nortel didn't have, or was this related to the timing, where Nortel was already in the gutter and RIM was still flying high?
Timing related. Nortel over-extended themselves in the 1990s by making a bunch of acquisitions related to networking. RIM could've bailed them out or acquired the company outright since they were very strong financially until just after 2008.
Remember, when the iPhone launched RIM was at their peak strength and Nortel was a forgotten husk.
Facebook was the one who started the OCP (Open Compute Platform) concept. Quite a number of companies have climbed on board. And there is a yearly conference in the March time frame in where quite a number of companies get together and show hardware based upon the open standards and concepts developed through OCP.
this is essentially an apache model or how it came to be. some of internet providers created and funded apache foundation to make a decent webserver that was flexible and reasonably fast, in terms of how fast a C based webserver can be.
I'm not too familiar with the design of apache but I believe nginx is also written in C. I'm curious about your comment because generally C is regarded as being the wrong choice for any given project for several reasons but speed is not one of them.
Are there other languages that are enabling faster web servers?
Actually what's stopping switch and router configuration from being something a local shop does for other local shops? At this point both the software and hardware aspects appear to be generic enough to be do-able. It also buys a local company really good local service. (Or,at least the potential for it.)
My thinking is if many companies can do it internally above a large enough scale then it may be a service opportunity for smaller entities.
> Plus, German telecom equipment maker ADVA Optical Networking is manufacturing the device and, as of a few weeks ago had nine customers trying it out for their telecom needs, a mix of big telecom companies and enterprises, it said.
That actually sounds a lot better, as it means Facebook isn’t actually manufacturing it, and it’s harder for a security agency to inject into it (although there will obviously still be spies)
