4 ms·
Cisco is circling the drain because their biggest customers finally realized they can build better products than Cisco using open source tools and commodity har
by exelius 9y ago
Cisco is circling the drain because their biggest customers finally realized they can build better products than Cisco using open source tools and commodity hardware. The big telecoms are all dramatically reducing spend and moving to proprietary in-house platforms.
Companies like Arista are selling white-box network gear at low margins because they know the market is commoditizing. Cisco got addicted to fat margins, but they have almost no patent advantage in most of their biggest markets. It gets hard to justify those margins after a while.
- microcolonel 9y agoI'll go one further and say that the white-box gear is better than the Cisco gear at this point. Had to deal with a network gateway at work last year which a) did not correctly implement STP, which we needed to hound Cisco to eventually patch, and b) hideously mistreated TCP connections so that they would be stuck open. I dealt with that second issue every day until my last at the company, I suspect that it's still b0rked if they haven't replaced it.
- myrandomcomment 9y agoAccton and Detla and Quanta are whitebox. Arista uses off the shelf switch ASIC from originally Fulcrum then Dune, Broadcom and now Xpilant. The motherboards and all other hardware is in house. Not whitbox.
- tptacek 9y agoI'd love it if someone could expand on this a bit. My infrastructure knowledge sort of dies around 2005-2006, when I left Arbor Networks, at a job where we had to know the details of the various TCAMs in different line cards for Cisco Cat6k switches. From '05-'06 I did a fair bit of experimenting with network processors (oddball embedded CPUs with fast packet processing mechanisms). Then I got sucked up into application security. The thing is, I've spent a career writing C code to parse, break up, and route packets for every conceivable protocol, in userland and several different kernels. What I can't tell is: is my skillset here totally outmoded because of the new network infrastructure model, or are we in a new golden age for it? (I am fine with either answer! I just want to know.)
- late2part 9y agoYou're fine but more real networking is migrating to sw stuff on aws+etc clouds. (Note: not sdn, think old school Zeus, Nginx, haproxy and orthogonally c* and kafka ) Moore's law etc makes what used to be done in fpgas and asics, now feasible in clusters of VMs. Yes there are exceptions but not that matter for masses.
- umbs 9y agoI can attest to OPs observation. Earlier in the day, I left detailed reasons for Ericsson's decline and laying off 25000 people. Main observation is: Networking is getting commoditized. I too have spent whole career writing C code and don't think there's a new network infra model. In past 25-30 years, it's matured enough to satisfy most of customer needs. I don't see growth here.
- yusyusyus 9y agoA big shining place is in DDoS mitigation. We demo'ed some in-house built replacements for $VENDOR and found it to be cost-effective to do some more easily implemented parts in-house. Based on some of my discussions with folks, I think people overestimate the cost-effectiveness of not doing some HW offload/out of kernel packet path stuff in specialized situations. Anyway, Amazon is looking for folks to work on their in-house DDoS mitigation pipeline per the reqs that float my way.
- exelius 9y agoI'd say we're in a new golden age, after which we'll have a bunch of generic, open source implementations with very little need to engage in that type of development again. This is actually happening all over the tech world -- companies like Google and Amazon are intentionally commoditizing key technology in their supply chain where there are a few powerful vendors extracting rents from everyone (see Cisco in the late 90s/early 00s). Just remember that 50 years ago, Ethernet didn't exist as a concept. This stuff moves fast, so the fact that your skillset has been useful for 40+ years tells me that it's probably not going to be highly valued for too much longer :)
- whatupmd 9y agoStuck at 1GbE/10GbE and yea commodity will catch up. When $VENDOR is using off-the-shelf chip to do same speed for cheaper they expose themselves to whitebox vendors competition. If there is no market or government pressure to level-up infrastructure (just optimizing existing network) not sure how $VENDOR will carve out any new revenue.
- scurvy 9y agoWhite box silicon is OK if you're deploying a campus, deploying a cloud where customers have no insight into the network, or situations where small packet performance doesn't matter. Everyone uses Broadcom (even though Mellanox spectrum is better), and Broadcom chipsets are trash. They're basically dump trucks that move lots of big packets, but can't do anything else. There will always be markets that want vendor designed NPU's because small packet performance, versatility, etc. Juniper QFX 10k has shown that they can compete and beat Broadcom at their own game. After watching Arista Broadcom switches drop traffic for years (they are lossy, sorry folks), I'm not touching them again. It's sad because they had something really special with their earlier products. I even owned some Arastra gear (yes, they renamed early on).
- scurvy 9y agoIt's not white box if it doesn't support ONIe. You can't replace the OS on an Arista switch. You're stuck with EOS.