|
Networks & Infrastructure
100 Gigabit Ethernet is on its way
2009 may be trial year, but deployment will be gradual
by Michael Kennedy
Juniper Networks’ recent announcement of a 100 Gigabit Ethernet (100 GE) interface for its T1600 Core Router has focused attention on the further scaling up of service provider networks. Though many other announcements of products and prototypes have been made over the last year, Juniper’s announcement is the first by a core router vendor and as such is an indication that major core networks are on the verge of moving to 100 GE backbones. In the same announcement, Verizon indicated that it is targeting commercial deployment next year.
The breadth of announced trials, designs, and prototypes indicates that 100 GE is receiving sufficient attention to be ready for commercial deployment soon. Huawei and Nortel have announced Ethernet switching projects, Huawei and Infinera have conducted trials and prototypes for transport solutions, and a number of chip and component announcements have been made including ones from Altera, NetLogic, Inphi, Gennum, Xilinx, and Alcatel-Lucent. Also Corning has announced 100 GE ready fiber and Ixia has announced 100 GE testing solutions. An IEEE 100 GE standard, 802.3ba, is expected soon.
Movement to 100 GE is driven by supply side and demand side factors. On the supply side enterprise Ethernet switches now routinely feature 10 GE uplinks while switches and routers designed for service provider aggregation networks also employ 10 GE ports. It therefore follows that data centers and service provider core networks which act as hubs for enterprise networks and service provider aggregation networks should provide backbones greater than 10 GE. OC-768 (40 Gbps) has been deployed to meet this need in service provider core networks. However, the 100 Gigabit Ethernet interface offers a lower cost solution and one that is compatible with the Ethernet technology used elsewhere in the network.
On the demand side, consumer, enterprise and mobile network traffic demand is expanding to the point where 100 GE will soon be needed to meet capacity requirements. Consumer use of high-speed Internet service (broadband) will be the largest and most immediate demand-side driver for 100 GE service. Video streaming is becoming an increasingly larger component of consumer Internet use. Cisco’s “Visual Networking Index” provides a highly detailed projection of global traffic growth with video traffic identified as a primary demand driver. The subscription-based Video on Demand (VoD) services offered by cable and telecom providers also are the primary demand drivers within service provider aggregation networks. My own traffic modeling efforts project tens of Gbps traffic requirements within each metro area. Note however that VoD traffic does not create too much core network traffic because the video servers are generally located within the metro network.
Wireless broadband including 3G and 4G also will contribute to rapid traffic growth. Wireless data traffic is the fastest growing market segment. The demand drivers are similar to those for wireline broadband traffic. However, wireless data is starting from a very small base so it will be several years before wireless data passes wireline data. Deployment of LTE beginning in 2010 and large reductions in the price of wireless data service will be needed to sustain traffic growth.
Enterprise networking also will add to total traffic growth. First, enterprises will be the source of most of the traffic that is downloaded to consumers. Secondly, the migration of enterprise private networking traffic from Frame Relay to MPLS VPN is an important demand driver. Unlike the consumer sector few new customers enter the market. However, traffic volumes from existing customers are exploding as they move from kilobit per second Frame Relay to megabit per second MPLS VPN. Data center applications such as cloud computing and software as a service (SaaS) also will require higher speed interfaces.
Though I have advanced arguments for the inevitable move to 100 GE it will be a long time before 100 GE is commonplace. As one data point I recently helped an enterprise client solicit bids for Carrier Ethernet services for several first and second tier cities in the southeastern U.S. Though responses were received from the seven largest service providers in the region no bidder was able to provide native Ethernet service and no 10 GE service was available even as trunks for the service providers’ own networks. In the absence of 10 GE aggregation networks 100 GE in the core network is overkill.
2009 appears to be the year for prototypes and trials for 100 GE with 2010 seeing the initial introduction of commercial services. Deployment will be gradual stretching out for many years.
Despite likely gradual deployment of 100 GE, large systems vendors must develop it now as a matter of survival. Even though there is a lot of product life left for 1 GE and 10 GE technologies, service providers take a very long view when choosing vendors of network infrastructure. Vendors must demonstrate a long term commitment to new technology such as 100 GE in order to prove that they have both staying power and vision. Consequently, sales of 10 GE today depend upon a convincing demonstration of the ability to deliver 100 GE in the future.
About the author: Michael Kennedy is co-founder and Managing Partner of Network Strategy Partners, LLC (NSP) www.nspllc.com —management consultants to the networking industry. He can be reached at mkennedy@nspllc.com.
|
