Debunking 10GBase-T Myths

Another major catalyst to 10GBase-T adoption will be the introduction of LAN-on-motherboard (LoM) chips. Expected in mid-2012, these devices will allow server manufacturers to also implement auto-negotiation technology into their gear. The implications of this development are quite profound. For the first time servers will come preconfigured with Ethernet connections able to negotiate to 100-Mbit/sec, 1-Gbit/sec or 10-Gbit/sec speeds depending on the capabilities of other devices in the network. The data center manager will want to be ready for this development by deploying 10GBase-T-capable switches that can extract the full capability of the server to which it is connected.

10GBase-T PHY power consumption is well managed

10GBase-T device power dissipation has been closely scrutinized and has been rapidly declining since IEEE 802.3 standardized the technology in 2006. Early PHY implementations were created using a 130-nm lithography manufacturing process, and they dissipated approximately 10W per port. By comparison, the 40-nm devices available today are capable of less than 4W-per-port dissipation. And the 28-nm devices anticipated in 2013 are expected to dissipate less than 2.5W per port. Figure 2 highlights this trend.

Figure 2:  10GBASE-T port power dissipation versus technology

Two protocols can further improve power dissipation

In addition to the reductions afforded by advances in semiconductor technology, Base-T systems—and 10GBase-T in particular—are able to take advantage of some unique and standards-based algorithms that exploit the nature of computer traffic to further reduce power dissipation. In particular, there are opportunities to improve efficiency when network equipment is idle for both sustained, and very short, time periods.

Wake-on-LAN (WoL) is a new networking standard formed by the Advanced Manageability Alliance whereby network equipment, such as a server, is put in sleep mode until awakened by a special network signal called a “magic packet.” The server’s NIC reverts to a very low power-dissipation mode during the sleep period, but remains alert and waiting for the magic packet. Once the packet arrives, the server is awakened and normal operation is resumed. Because the wakeup time associated with WoL is typically tens of seconds, this power-management strategy is best suited for use when servers are expected to be idle for long periods of time, such as at night or during other lengthy periods of inactivity. Even the most active of data centers experiences periods of time in which only a portion of its capacity is needed. This is a natural consequence of over-building resources to accommodate peak compute demands and the temporal and seasonal fluctuation in those demands due to non-uniform user locations and time schedules.

WoL can take advantage of these demand fluctuations with startling results; putting even a single server with a typical power dissipation of 500W to sleep gains much more benefit than the difference in power of hundreds of transceiver devices. Equally important, 10-Gbit/sec Ethernet deployed over optical media or SFP+ direct-attach assemblies is not designed to support the WoL protocol at this time and, as a result, these systems always dissipate their full power. WoL is an important strategy employed uniquely by 10GBase-T to reduce overall power consumption in the data center.

While WoL is designed for lengthy idle periods, another technology called Energy Efficient Ethernet (EEE) is specifically designed to take advantage of the bursty nature of computer traffic. It is the case that typical Ethernet traffic contains many gaps, which can range in duration from microseconds to milliseconds. Heretofore, these gaps have been filled with so-called “idle patterns” or waveforms containing no real computer information, but whose transitions can be used for maintaining clock synchronization between transceivers. The EEE algorithm exchanges those idle patterns for Low Power Idle (LPI) mode where very little power is dissipated.