Making a Rapid Move to 100GE
The 40GE/100GE standards are ready, but the industry has a long way still to go before we see high volumes. What is holding back adaption? Big content providers like Facebook and Google have been pushing for commercial viable 100GE systems for some time. The Internet backbone players are increasingly challenged to keep up with bandwidth demand. I believe Michael Howard of Infonetics summarized it well at the Ethernet Summit in San Jose recently. As he put it, “Early components are expensive as well as large, power-hungry and hot, and there are still several generations to go in downsizing these parts for more economical systems.“
So, let us look at an example of a 100GE board of today, using in-house designed packet processing silicon:
The line card is packed with silicon to perform packet processing, traffic management and buffering for 100 Gbit/s of traffic. Here is where the new member of the HX family, the HX336 comes into play. The HX336 includes a 100 Gbit/s traffic manager with deep packet buffering in off-chip DRAM. By utilizing the service density of the HX family, the line card above can be optmized to only use two chips for packet processing and traffic management, a pair of HX326 and HX336. The HX336 is used in the egress pass supporting both packet processing and traffic management. The HX326 is responsible for traffic classification and packet processing in the ingress path. Reducing the number of chips for packet processing, traffic management and buffering from four to six to two result in significant power and cost savings. Several of our customers have witnessed a 50% reduction of power consumption.
The new HX336 network processor is Xelerated’s contribution to the industry’s transition to 100GE and optical transport networks OTU4 standard. The transition may not be immediate, but as technologies matures, the shift may be stronger than expected.
| Comment





