A new task force and others within the IEEE community are working on new specifications to both expand the set of current 40 and 100 Gigabit Ethernet specifications as well as determine what should come next.
At the 40 and 100 Gigabit Ethernet rates, the P802.3bm task force currently is examining ways to create a “step-function improvement” in 40 and 100 Gigabit Ethernet technology costs as well as fill a hole in the 100 Gigabit Ethernet coverage area. The new specifications cover both multimode and singlemode fiber applications. They include:
- 100GBase-SR4, a 100-Gbps PHY for operation up to at least 100 m of multimode fiber. The PHY will be based on 4x25-Gbps parallel optics housed in a QSFP+ optical transceiver module.
- potentially another 100-Gbps PHY for multimode fiber under the 100GBase-SR4 effort, this one for lengths of at least 20 m; once again, it likely will be based on 4x25G parallel optics in a QSFP+ form factor
- 100GBase-nR4, a 100-Gbps PHY for applications up to at least 500 m of singlemode fiber. This specification appears to be a direct response to the concerns of major data center operators, such as Google and Facebook, that led to the creation of the 10X10 MSA (see “Google joins tech vendors in 10x10G 100 Gbps optical transceiver multi-source agreement”).
- 40GBase-ER4, a 40-Gbps PHY for applications of at least 40 km of singlemode fiber for metro applications
- a new 4x25-Gbps CAUI electrical interface.
As noted above, the technology that will underpin the multimode fiber specifications has already been determined, but the task force remains undecided about the singlemode technology foundations. Three proposals are being considered:
- parallel optics (which proponents are calling 100GBase-PSM4)
- CWDM
- complex modulation, specifically PAM.
The three options were debated during a task force meeting in Geneva, Switzerland at the end of September. Speaking before the meeting, a source within the Ethernet Alliance suggested that the CWDM approach has comparatively few proponents, while Cisco appears to favor the PAM option. It likely will take six months for the task force to settle on an approach, the source predicted. The finished specifications likely won’t be ratified until March 2015.
Meanwhile, the IEEE has already begun to consider what comes next – and is looking for help. As its name implies, the new Industry Connections Higher Speed Ethernet Consensus group is tasked with developing industry agreement on the IEEE’s next direction (see “New IEEE 802.3 group seeks consensus for next-generation Ethernet”). John D’Ambrosia of Dell chairs the group, whose website is http://www.ieee802.org/3/ad_hoc/bwa/index.html.
Ethernet specifications generally have jumped by 10X, which would imply the next data rate milepost will be 1 Tbps. However, even with the consensus group just getting started, there is growing sentiment that 400 Gbps is the next likely step. As stated in a slide David Ofelt, a distinguished engineer at Juniper Networks, presented at an Ethernet Alliance session at ECOC in Amsterdam last month, “Reality is that 1TbE is currently impractical.”
In a different presentation at ECOC, Jeffrey Maki, another distinguished engineer at Juniper, speculated on what 400-Gbps modules might look like. The form factors could be similar to those now offered and contemplated for 100 Gbps – namely, CFP, CFP2, and CFP4 – but with different names, Maki believes. (The “C” would be replaced by “CD” to create “CDFP,” etc.) The electrical interfaces would be quite different, however. Maki offered 16x25G for the CDFP, with 8x50G being necessary for a CDFP2 module, and 4x100G to accommodate the small size of a CDFP4 module. He believes a first-generation CDFP optical transceiver would need to cost no more than four 100-Gbps CFPs and consume no more than 24 W of power.
While 40 and 100 Gigabit Ethernet technology is only now being fielded, there’s enough feedback from the field to spark a new generation of such standards. Meanwhile, given the fact that rapid growth in bandwidth is expected to continue into the foreseeable future, work on the successor to 100 Gigabit Ethernet as the top available data rate should begin relatively soon.