by Jim Duffy
So with word circulating around the industry of a possible Cisco spin-in developing a cloudy switch for software-defined networking (SDN) and/or distributed data storage, what's the status of Cisco's "Jawbreaker" fabric switching project?
Sources in the fabric and SDN industry say Jawbreaker, which was to be a merchant silicon-based response to Juniper's QFabric, has either been:
• Refocused toward the enterprise campus as a Catalyst switch adjunct;
• Repositioned into Insieme, the not-so-stealthy startup spin-in run by Cisco's top three engineers -- Mario Mazzola, Luca Cafiero and Prem Jain;
• Or killed entirely due to the apparent challenges Juniper's facing with QFabric. QFabric's lengthy trial and ramp cycle no longer requires a competitive marketing response from Cisco, some sources believe.
Cisco has never publicly acknowledged the existence of the Jawbreaker project, which came to light 13 months ago, so information from sources couldn't be confirmed. A year ago, sources said Jawbreaker was based on Broadcom's Trident+ chipset and had two components: a fabric core with 40G interfaces and a 48-port 10G switch with 40G uplinks. It was to ship later this year, sources said at the time.
It still might. But now that Cisco's rolled out 40G Ethernet modules for the Nexus 7000 and perhaps some uplinks soon for the Nexus 5000s, as well as extending Nexus Fabric Extenders, it at least appears that Nexus and FabricPath are fairly well along in the data center fabric investment and development cycle.
As a result, Jawbreaker has been retargeted toward the enterprise, some fabric sources believe. Its application there could be a higher-speed, lower-latency campuswide fabric as an adjunct or upgrade to the Catalyst installed base.
"It's being done by the Catalyst group," one fabric source said. "It may be refocused on campus only. But as a data center fabric, the rumor mill seems to be losing some steam."
Another option for Jawbreaker could be as a switch/controller for SDNs, especially now that rival HP is focusing OpenFlow on the enterprise with the recent introduction of 16 OpenFlow-enabled switches. OpenFlow/SDN control of merchant silicon-based hardware is said to be a threat to Cisco's custom ASIC/tightly coupled software-based switch dominance in enterprises.
And though Jawbreaker is said to be based on merchant silicon, there's a component sources say wasn't: the fabric core, which one SDN source called a "parent" switch. The parent was to be a typical Cisco development of a platform based on custom ASICs tightly coupled with the switch software.
This Jawbreaker parent was under the purview of Cisco engineers Tom Edsall and Dinesh Dutt until both jumped over to Insieme, the distributed data storage/OpenStack/SDN(?) startup said to be negotiating a spin-in deal with Cisco, the SDN source says. Now the Jawbreaker parent may be at Insieme because its parents are.
"There are two pieces to this strategy," the SDN source says. "The first was getting a merchant silicon based platform out the door based on Trident+ from Broadcom. The second was a 'parent' switch. This second piece is exactly what Tom and Dinesh were working on at Cisco before they left."
Edsall and Dutt's involvement with Insieme could not be confirmed. But another fabric source says the engineers could have left Jawbreaker behind if they jumped to Insieme.
This source says he believes Jawbreaker has been canceled entirely since QFabric, its market quarry, has been hobbled by a purported slow ramp in sales. Juniper says QFabric is exceeding market expectations.
And since Jawbreaker was largely based on merchant silicon and likely less expensive to make, it was just as inexpensive -- for Cisco, anyway -- to kill it. Marketing mockups are usually one and done projects.
"If they will deliver this at this point or use this technology as the basis for what Mario and Co. are doing is not clear," the SDN source says.
"Cisco can afford to have multiple projects that overlap and not necessarily bring all of them to market," a fabric source says.
So with word circulating around the industry of a possible Cisco spin-in developing a cloudy switch for software-defined networking (SDN) and/or distributed data storage, what's the status of Cisco's "Jawbreaker" fabric switching project?
Sources in the fabric and SDN industry say Jawbreaker, which was to be a merchant silicon-based response to Juniper's QFabric, has either been:
• Refocused toward the enterprise campus as a Catalyst switch adjunct;
• Repositioned into Insieme, the not-so-stealthy startup spin-in run by Cisco's top three engineers -- Mario Mazzola, Luca Cafiero and Prem Jain;
• Or killed entirely due to the apparent challenges Juniper's facing with QFabric. QFabric's lengthy trial and ramp cycle no longer requires a competitive marketing response from Cisco, some sources believe.
Cisco has never publicly acknowledged the existence of the Jawbreaker project, which came to light 13 months ago, so information from sources couldn't be confirmed. A year ago, sources said Jawbreaker was based on Broadcom's Trident+ chipset and had two components: a fabric core with 40G interfaces and a 48-port 10G switch with 40G uplinks. It was to ship later this year, sources said at the time.
It still might. But now that Cisco's rolled out 40G Ethernet modules for the Nexus 7000 and perhaps some uplinks soon for the Nexus 5000s, as well as extending Nexus Fabric Extenders, it at least appears that Nexus and FabricPath are fairly well along in the data center fabric investment and development cycle.
As a result, Jawbreaker has been retargeted toward the enterprise, some fabric sources believe. Its application there could be a higher-speed, lower-latency campuswide fabric as an adjunct or upgrade to the Catalyst installed base.
"It's being done by the Catalyst group," one fabric source said. "It may be refocused on campus only. But as a data center fabric, the rumor mill seems to be losing some steam."
Another option for Jawbreaker could be as a switch/controller for SDNs, especially now that rival HP is focusing OpenFlow on the enterprise with the recent introduction of 16 OpenFlow-enabled switches. OpenFlow/SDN control of merchant silicon-based hardware is said to be a threat to Cisco's custom ASIC/tightly coupled software-based switch dominance in enterprises.
And though Jawbreaker is said to be based on merchant silicon, there's a component sources say wasn't: the fabric core, which one SDN source called a "parent" switch. The parent was to be a typical Cisco development of a platform based on custom ASICs tightly coupled with the switch software.
This Jawbreaker parent was under the purview of Cisco engineers Tom Edsall and Dinesh Dutt until both jumped over to Insieme, the distributed data storage/OpenStack/SDN(?) startup said to be negotiating a spin-in deal with Cisco, the SDN source says. Now the Jawbreaker parent may be at Insieme because its parents are.
"There are two pieces to this strategy," the SDN source says. "The first was getting a merchant silicon based platform out the door based on Trident+ from Broadcom. The second was a 'parent' switch. This second piece is exactly what Tom and Dinesh were working on at Cisco before they left."
Edsall and Dutt's involvement with Insieme could not be confirmed. But another fabric source says the engineers could have left Jawbreaker behind if they jumped to Insieme.
This source says he believes Jawbreaker has been canceled entirely since QFabric, its market quarry, has been hobbled by a purported slow ramp in sales. Juniper says QFabric is exceeding market expectations.
And since Jawbreaker was largely based on merchant silicon and likely less expensive to make, it was just as inexpensive -- for Cisco, anyway -- to kill it. Marketing mockups are usually one and done projects.
"If they will deliver this at this point or use this technology as the basis for what Mario and Co. are doing is not clear," the SDN source says.
"Cisco can afford to have multiple projects that overlap and not necessarily bring all of them to market," a fabric source says.
No comments:
Post a Comment