Standards-based 40- and 100-gigabit Ethernet switches and routers are starting to show up in enterprise networks, and it’s easy to understand why – fast downlinks require faster uplinks. The current solution of aggregating multiple 10G links works well but only scales up to a point.
There are a number of factors of driving the adoption of 40/100G in enterprise networks. Servers for some high-performance application now use 10G interface cards requiring a faster uplink at the switch. It won’t be long before 10G interfaces will be a standard part of server motherboard, just as gigabit Ethernet comes standard today.
“Future growth in the Ethernet switch market is expected to come primarily from the data centre, with compute changed such as server/storage consolidation, virtualisation and fabric infrastructure increasingly demanding a change in the way data centre networks are designed, and also greater bandwidth,” says Severine Real, Senior Principal Research Analyst, Gartner.
Data centre is expected to be the primary market for 40GB with 100 GB finding its place core network aggregation, mainly in carrier networks. “Data centres need to need to go beyond the redundancy requirements of yesterday to a more future-proofed infrastructure in order to support the demanding availability requirements of today’s applications. This requires organizations to support new and upcoming technologies like 40G/100G Ethernet and standards, and also to choose solutions that will provide an open and flexible architecture to support the evolving needs of businesses,” says Sakkeer Hussain, Sales & Marketing Manager at D-Link MEA.
Markus Nispel , Chief Technology Strategist at Enterasys agrees that for the enterprises it will be clearly the data centre. “Service providers might require it for core aggregation but in the enterprise core 40G and 100G won’t play a significant role initially.”
Vendors also say that the market for 40G will grow faster than 100G as the latter is not cost-effective now. “Both 40G and 100G have their place in the network, as needs of every individual network operator are unique. This said, one of the strong determinants of growth in both 40G and 100G is pricing, and as volumes of 100G systems sold worldwide increase (with time), one can imagine it will also become more economical for network operators to deploy this higher-capacity option,” says Saad Khan, Regional Managing Director of Ciena.
Ashish Saxena, Solution Architect, Alcatel-Lucent Enterprise, adds that this would depend on the services and applications as well as the needs of various enterprises. “Currently, we are seeing a growing demand for 40G. 100G is already in discussions and is an aspiration. Network convergence and new devices on the network will continue to grow and the demand for bandwidth will continue to grow too.”
Steven Huang, Director of Solutions and Marketing, Huawei, says demand in data centres has mainly been for 40G up until now in the region, although we are seeing 100G per second links between large switch and router points beginning to emerge in data centres.
“As network architecture evolves, internet traffic continues to grow and advance, giving room for greater energy efficiency and making faster connectivity more affordable for businesses in terms of power consumption per traffic load. The demand for 100G will grow as the industry is already paving the way for 400G to a terabit and beyond,” he adds.
Currently the cost of a 40G Ethernet link on single mode or multimode fibre is roughly about four to five times the amount of a 10G link, and a 100G Ethernet link can cost up to twenty times the amount of a 10G Ethernet interface. “Such economics, coupled with the fact that the IEEE 802.3ba standard for 40/100G Ethernet is still three to four months from approval, is keeping the demand subdued even though the interest in these technologies is relatively high. Economic studies based on this say that 40G can expect a similar price decline curve as 10G when it ships in increasing volumes, whereas 100G is expected to follow the decline of 40G,” says Hussein.
Gartner hasn’t published a forecast for 40/100G yet but it plans to do so later this year, as it is still a very small market. “For the market for 40/100G switches to start off, prices will have to decline to be more enticing, and cost effective. Otherwise, enterprises will keep trunking 10G links together to avoid the costly expense of bandwidth.”
Though there is no doubt the adoption of high-speed Ethernet will help to push the prices points of 10G down, Nispel from Enterasys says it will take another round of switch architecture before the 10G pricing for high-end switches and routers will come down significantly.
What changes do companies require to do in order to migrate to 40/100G? Saxena from Alcatel-Lucent says companies must plan for the long term today. “The investments that are made today must be well protected for the future migrations to QSFP+ transceivers for 40G, making migrations to 40G/100G just a minor hardware change with bare minimum investment. At Alcatel-Lucent, we have designed our switches with future readiness for 40G/100G and with minor hardware changes.”
Samer Ismair, Systems Engineer at Brocade, adds that as with any new generation of technology, one design goal was to leverage as much existing technology as possible. By minimizing the number of new interfaces, the interfaces become less expensive and take advantage of volume production and simplicity. To meet this design goal, only three media modules will be used in the first generation of 40 Gigabit and 100 Gigabit Ethernet: QSFP, CXP, and CFP.
While many enterprises may have to redesign or optimise their networks to migrate, they will also have to look at the physical infrastructure layer. “The 100G standard was ratified in 2010 by the IEEE 802.3ba committee for transmissions up to 150 meters of OM4 fiber cable. The common distances for both 40 and 100G rates are the result of using inverse multiplexing over parallel fibres, wherein the signals are transmitted over multiple lanes, each operating at 10G. The 40G interface employs four of these lanes in each direction, while the 100Gb/s interface employs ten,” says Ciaran Forde, VP of CommScope.
Migrating to high-speed Ethernet will certainly require changes. The good news is that the changes are evolutionary in nature, and can be rolled out incrementally. Ethernet has undergone many changes over the years, but the traffic on high-speed Ethernet networks retains the same basic characteristics as previous generations. In the end, it really is “just Ethernet.”