Is 10GBASE-T widely available on servers?
10GBASE-T has been available in servers as an add-in card since 2008. Wide availability was not immediate since the introduction of new technology onto server motherboards is driven by the new platform design cadence of approximately 2 years. Maturing 10GBASE-T products intersected with new server platforms in 2012 when 10GBASE-T Ethernet ports became available either as LAN on Motherboard (LOM) or a configurable daughter card from many server vendors in 2012, resulting in over a half million 10GBASE-T server ports shipped that year. The trend continues as additional products are coming to market in 2013, and shipment of 10GBASE-T ports are projected to outnumber SFP+ ports in 2015 (Source: Crehan Research).
Is 10GBASE-T widely available on switches?
10GBASE-T has been available on switches since 2008, and vendors have various 10GBASE-T switches solutions.
10GBASE-T adoption has been fueled by a wide ecosystem of adapter/switches vendors coinciding with the release of the 40nm 10GBASE-T PHY in 2012 with new innovations, cost points and technology maturity driving large deployment.
Deployments have been in Data Centers mainly, although solutions are also available for Campus. Shipment of 10GBASE-T ports are projected to outnumber SFP+ ports in 2015 (Source: Crehan Research).
Is 10GBASE-T ready for prime time?
10GBASE-T is ready for prime time with the latest proof points showing adoption growth, and a large ecosystem of server and switch vendors.
Articles on adoption:
- Tips & Trends: 10GBASE-T adoption status and forecast, Debunking the 10GBASE-T Myths
- How is 10GBASE-T Being Adopted and Deployed
What are the drivers behind 10GBASE-T adoption?
10GBASE-T has been widely adopted since 2012 and the trend continues. The early drivers were simplicity and reuse of existing structured cabling with ease of migration. Nowadays, cost efficiency is becoming more and more relevant.
Major drivers are listed below:
- Cost effective at up to 100m distance
- 10GBASE-T Server LOM
- Ease and reuse of standard/structured cabling
- Overall migration towards 10G Ethernet
- Ease of 1GBASE-T to 10GBASE-T migration
- Multi-speed flexibility: Platforms support dual speed (1000BASE-T/10GBASE-T) or triple speeds (100BASE-T/1000BASE-T/10GBASE-T)
- Architecture Flexibility: Supports Top of Rack, Middle of Row or End of Row architectures
What is the status of EEE? How does that help with 10GBASE-T?
Energy Efficient Ethernet is defined by the IEEE as 802.3-az. EEE enables power reduction in the PHY during periods when no data is being transmitted. The policy for entering the Low Power Idle (LPI) mode is set by the higher layers, with the decision to enter LPI often based on meeting a criteria of exceeding some pre-set limit of consecutive idles. Some have argued that EEE is not important for 10GBASE-T, after all, most applications in the data center are “always on” type applications. However Ethernet by its very nature is a bursty packet based protocol. When the criteria to enter the LPI mode is set at the minimum length, the amount of time spent in the lower power state is maximized, and saving power is quite important. The savings will be a function of the application, but can be significant. The major downside is the latency when exiting LPI of 4 to 7 microseconds.
A number of applications are in a position to gain a major benefit from EEE. Not all 10GBASE-T applications have symmetrical data flow. The nature of EEE in 10GBASE-T is that it applies independently in each direction. So a link with high traffic in one direction, and light traffic in the opposite direction can leverage EEE to minimize overall power. Also, systems entering standby modes can maintain their network connections at the lower power permitted by EEE, speeding the recovery of the network connection upon exiting standby mode.
What’s beyond 10GBASE-T?
Our industry has delivered successive generations of faster speeds over twisted pair cabling. We went from 10, to 100, to 1000BASE-T and then to 10GBASE-T, all of which were seen as significant achievements in their time. It was perceived that twisted pair could not support such high data rates. We’re now poised to attack the next increment in data rates on twisted pair, 40GBASE-T.
The challenge has been taken by the IEEE, with its Next-Generation BASE-T Study Group in conjunction with ISO and TIA cabling standards bodies who are developing standards for cabling to support the application. This group has had presentations that demonstrate that 40GBASE-T is feasible in advanced CMOS technologies, given currently known IC design techniques and twisted pair cabling technologies.
Why would we want to do this, when existing QSFP Direct Attach Copper is available to provide 40GE connections today? The answer is in what twisted pair cabling can do what DAC cannot. Twisted pair cabling can support links up to 30 meters of structured cabling for 40GBASE-T, which lets one interconnect support both Top of Rack (point to point cabling) as well as End of Row or networking isle topologies (structured cabling thru patch panels). With backward compatibility to previous generations of BASE-T, 40GBASE-T will easily interoperate with 10GBASE-T and 1000BASE-T, and will offer the customary 48-port densities in a 1RU form factor.
The IEEE Study Group chose the 30m minimum distance with practicality in mind. 30m was seen as sufficient to support the targeted application in the data center. The additional benefit of limiting the distance was to minimize the implementation power and complexity, and thereby speed market adoption of this new technology.
Can 10GBASE-T support PoE?
During the Call-For-Interest presentation, it was noted that increasing bandwidth requirements of wireless access points is driving the demand for Power over Ethernet to provide support beyond 1000BASE-T. The IEEE 802.3 4 Pair Power-Over-Ethernet Study Group met for the first time at the IEEE 802.3 May Interim Meeting and will begin the work of determining the project objectives. It is anticipated that PoE Support for 10GBASE-T will be considered by this group as a possible objective.
More information at http://standards.ieee.org/news/2013/4pair_poe.html