re: Cisco Unveils the HFRYour right...this is the most significant press release in Cisco's history...ah yeah...they finally admitted that IOS and GSR are out of date...
And here's what they'll use to fix it...mystery meat...
...performance and density that match Junipers...and a platform that scales to 92Tb...is just a 'me to' and fantasy marketing...
All old news...is boring quite frankly...C'mon Cisco...you used to lead the way...what's next...
I think most people know about the need to halve the total throughput calculation. But what about the backplane speed? (--which is probably irrelevant for non-blocking switches but means everything to blocking ones.)
The reason question is that the stated backplane speed of a Cisco 6500 (Sup 1) is FOUR times the actual full duplex speed--at least according to a 1999 Cisco whitepaper. Here, even one way traffic from one port to another is counted twice (as one ingress and one egress); hence, the full duplex traffic is counted four times. In Sup1's case, the stated backplane speed of 32 Gbps is just 8 Gbps!?
Does it mean we divide the throughput rate by two but the backplane speed by four? (Is it another industry-wide practice or just Cisco-speak?)
re: Cisco Unveils the HFRWhile I do not have the first-hand experience with HFR, from CSCO papers it is clear that at least four new products are introduced with CRS-1
- new modular and distributed OS - new 40G ASIC - new full-rack chassis - new optical backplane interconnect
It was pointed by several people that optical interconnects are expensive and are only compelling to few customers requiring extreme densities; most others will be better off with the single-chassis designs and the 5-year forklift cycle. If true, this effectively limits the real-world CRS-1 appeal to the customers in need of 33 to 64 OC-192 ports per system (possibly extended to 128 if a cheap back-to-back is possible). Quite obviously, as of today this sweetspot is fairly small compared to the range of 1-32 ports offered by JNPR and Procket.
Interestingly enough, almost no one commented on the new forwarding ASIC, which is described to be fully programmable. This means it can probably power a broad range of the follow up edge/core products, starting from half rack and down to fabricless 1RU blades
So if Cisco is serious about the whole development effort, I would expect to see many product launches in the next couple of years This also dooms the current 10K/12K/6K lineup
re: Cisco Unveils the HFRHeard from three different people in the last two days. They are in serious discussions, due diligence phase. Bobby will probably retire as there's no Vice Chair slot, no President slot unless they want Dolce to bolt. But maybe he won't retire after all. Some hangups there. Not a done deal but about 80%. Anybody got any G2?
re: Cisco Unveils the HFR"Nothing is cheap which is superfluous, for what one does not need, is dear at a penny. " - Plutarch
That's a lot of digesting in six months. And one way to get folks wondering about where the future is bringing them. Can't imagine the resulting dilution would do much for the stock options of the rank and file, either (bad enough after Netscreen). I hope it wasn't Foundry's stock price which became the ultimate seducer.
What is it that Juniper needs so badly? Can't they develop an enterprise or metro-ethernet product with the Netscreen product and routing expertise of Juniper on their own? I've got to imagine that it's easier to put the routing/switching into the security realm than the other way around.
Where's fw23 to shed some LIGHT on the situation?...
re: Cisco Unveils the HFRThis would be very interesting in light of Foundry's serious ongoing discussions with Juniper. Hopefully Procket will have a good exit and Foundry buying them would be a good thing. This would make Foundry a real technology company with a future. Some IP they can build around and a cool product. Gives them Cisco-like leverage and Juniper is the bigt loser in this scenario because they still don't have a portfolio for enterprise. Their routing program's been a disaster. So buying Procket enables Bobby to give Juniper a run, as opposed to getting bought by them, to work for Scott (Jim), (Marcel) or retire, admitting defeat (would Bobby succomb that easily?) I see them taking the money (Juniper buyout) and running.
And here's what they'll use to fix it...mystery meat...
...performance and density that match Junipers...and a platform that scales to 92Tb...is just a 'me to' and fantasy marketing...
All old news...is boring quite frankly...C'mon Cisco...you used to lead the way...what's next...