you trade a lot of latency to make GPU parallelism work for packet processing/classification. some massively clever work around hiding it — but simply no way to avoid it. thus its a niche solution
984 karma · joined February 19, 2013
dropping a packet after processing overhead and/or explicit classification work is a useful benchmark yes
things like P4 will move the competitive advantages farther up the stack where they belong
Not great. And it's beyond disingenuous to call peering "free". How does DE-CIX make any money? Do your line cards and optics and cross connects get donated?
you have to wonder, if Berlin is just a bunch of nonsensical leftists and anarchists, why would Google be so interested in opening an engineering office here? it couldn't have anything to do with the talent they want to hire and retain wanting to live in Berlin, would it?
I mean, does Google really give a shit about 18euro pqm versus 24 o4 25 in Frankfurt? highly doubtful. so what do you think the reason is, if you truly believe your Berlin anarchy thesis?
Believe it. HE has many, many netiron-based routers in their network and many, many ASNs as transit customers. the hardware doesn't have the capacity to implement complete filtering for everyone.
between hardware-constrained equipment being replaced and the tireless work Job Snijders (& others) are doing with route security, these kinds of episodes will become far more difficult and occur much, much less.
that's just the public stuff!
galois fields are very interesting for random linear network coding, so of course there is a very thorough patent group (recently sold off, I believe) around it.
given the seriously transformative effects it can have on communications networks -- especially wireless -- I'm both saddened and unsurprised it's in basically nothing as of yet. because it's patented to Hell & Back.
I believe it's high time the long history of anticompetitive actions by Intel end, and their near/effective monopoly in major market segments be regulated.