New gigabit standard: How fast is fast enough?

Last June, the 10GBASE-T standard was published promising 10 Gbit operations on copper cabling. This standard was originally targeted toward data centers, but we have all heard the old adage "Today's server is tomorrow's desktop," which means that something faster is needed at the core. The Ethernet Alliance and IEEE are working on just such a project. Can you imagine having potentially 40 Gbit and 100 Gbit?

If we look for the next killer application, we don't in fact really know what it will be. Certainly, storage will benefit, as will throughput to servers. Companies that today are virtualizing servers will surely see an uptake in bigger pipes. Collaboration and video applications will benefit from larger throughput. Multiple dwelling units and service providers alike will see a big benefit. But what will the next speed be?

We certainly have WAN equivalents of both 40 Gbit and 100 Gbit. The key determination here may very well be answered in the medium of transport -- having copper category or fiber either single-mode or multimode. When standards are developed, they go through a very consistent process. Individuals from companies that wish to bring forth the technology get together and provide a presentation on the merits of the idea. Although each participant in IEEE is officially an individual not representing company interests, initial calls for interest tend to be from individuals from complementary industries. Companies and other sponsors present information to IEEE to determine whether the project should move forward. If so, a study group is formed. The study group is responsible for writing the project authorization request and answering the five criteria. These are technical feasibility, economic feasibility, broad market appeal, unique solution, and interoperability with other standards. Once these have been completed by the study group, another vote is taken. If it is determined that the project is a go, a task force is formed of those members interested in participating in the formation of the standard. A few approvals later, a standard is born.

The higher-speed networking standard is in the study group stage. The study group-approved objectives so far are 100 Gbit on 50 micron laser optimized fiber or OM3 fiber (at least 100 meters) and single mode fiber (at least 10 km). The latter may change at some point to 40 km. There is also talk of having 40 Gbit per second. This may very well be the best alternative for copper but may provide an additional step for fiber applications between 10 Gbit and 100 Gbit. The higher-performing copper plants today, category 7/Class F (which is the only currently published standard ISO-IEC 11801 that supports 10GBASE-T IEEE 802.3an the category 6a standards both from ISO-IEC and TIA will publish this year) and the newer 7_A (which is a 1 GHz cabling system) have been modeled and shown to have the potential to support a 40 Gbit copper alternative. This may be a great application for storage and server connections in the data center. Even if a core is upgraded to 100 Gbit, the 40 Gbit for horizontal locations may prove to be popular until 100 Gbit is really needed.

It is estimated that it will be quite some time before a CPU can take full advantage of a 100 Gbit connection. Some estimates put this capability near the year 2015, while CPUs could take advantage of a 40 Gbit alternative closer to 2010. In comparison with wide area speeds, either alternative or even both would have some nice equivalents. Just think -- 10-15 years ago, we were on DOS and green-screen applications, and Windows was new. In the next few years, we could be moving data and carriers around the Internet at speeds 40-100 times faster than we can today. Makes you think of drinking from a fire hose today; but in the future, who knows?