40 Gigabit Ethernet QSFPs boost port density and reach

Avago Technologies and Reflex Photonics have announced extended reach 40 Gigabit Ethernet (GbE) QSFP+ transceivers.  As data centres grow in size, there is a need to link equipment over distances greater than 100m, as defined by the IEEE 40 Gigabit Ethernet standard.

"For the larger data centres being built today, reach is becoming more and more important"

I Hsing Tan, Avago 

Avago’s eSR4 QSFP+ transceiver extends the reach of 40GbE over multimode fibre beyond the IEEE 40GBASE-SR4 specification, to 300m over OM3 and 400m over OM4 multimode fibre.

Reflex Photonics’ 40GbE QSFP also achieves 300m over OM3 fibre and while it has not tested the transceiver over OM4 fibre, the company is using the same optics that it uses for its CFP which meets 450m over OM4.

“This [QSFP] is aimed at large data centres operated by the likes of a Google or a Facebook,” says Robert Coenen, vice president, sales and marketing at Reflex Photonics. Such data centres can have link requirements of 1000m. “The more reach you can give over multimode fibre, the more money they [data centre operators] can save.”

The eSR4, like Avago's already announced iSR4 (interoperable SR4) 40GbE QSFP+ transceiver, supports either 40GbE or four independent 10GbE channels. When used as a multichannel 10GbE interface, the QSFP+ can interface to various 10GbE form factors such as X2, XFP and SFP+, says Avago. 

The iSR4 also increases the faceplate port density of equipment from 48, 10 Gigabit Ethernet (GbE) SFP+ ports to up to 44 QSFP+ 40GbE ports. Avago says that one equipment vendor has already announced a card with 36 QSFP+ ports. The iSR4 QSFP+ also reduces the overall Gigabit/Watt power consumption to 37.5mW/Gbps compared to 100mW /Gbps for the SFP+.  The eSR4 has half the power consumption, which puts it around 50mW/Gbps.

But the iSR4 matches the reach of the IEEE 40GBASE-SR4 40GbE standard: 100m for OM3 and 150m for OM4-based fibre. "This [reduced reach at 40GbE] creates an issue for data centre operations," says I Hsing Tan, Ethernet segment marketing manager in the fiber optics product division at Avago. "They require additional investment to redo all the wiring in current 10GbE infrastructure to support a shorter reach." 

With the extended reach 40GbE QSFPs the reach associated with 10GbE interfaces on OM3 and OM4 multimode fibre is now restored.

The iSR4 module is available now, says Avago, while the eSR4 will be available from mid-2012. Reflex’s Coenen says it will have samples of its 40GbE QSFP, which also supports 40GbE and 4x10GbE,  by May 2012.

What has been done

For Avago's iSR4 QSFP+ to operate as four, 10GbE channels, it has to comply with the 10GBASE-SR optical standard. That is because 10GBASE-SR supports a maximum receive power of -1dBm whereas the 40GBASE-SR4 has a maximum output power of 2.4dBm. The transmitter power of the iSR4 has thus been reduced. "We force the output of the transmitter down to -1dBm," says Tan.

To achieve the greater reach, the eSR4 uses a VCSEL design with a tighter spectral width. Other parameters include the optical modulation amplitude power and the wavelength. These affect the resulting fibre dispersion. “Once you control the spectral width, you can design the other two to meet the specs," says Tan.  

The Avago 40GbE QSFP+ modules use an integrated 4- channel VCSEL array and a 4-channel photo-detector array.

Significance

The 40GbE short reach interfaces play an important role in the data centre. As servers move from using 1GbE to 10GbE interfaces, the uplink from aggregation 'top-of-rack' switches must also scale from 10GbE to higher speeds of 40GbE or 100GbE.

However existing 100GbE interfaces make use of the CFP module which is relatively large and expensive. And although the 100GbE standard has a clear roadmap leading to CFP2 and CFP4 modules, half and a quarter of the size of the CFP, respectively, these are not yet available. 

40GbE QSFP+ transceivers do exist and offer the equipment faceplate density improvement vendors want. 

The QSFP+ also benefits existing 10GbE designs by supporting nearly 4x the number of 10GbE on a card. Thus a new blade supporting up to 44, 40GbE QSFP+ transceivers can interface to up to 176 10GbE transceivers, a near fourfold capacity increase.

According to Avago, between 10% and 20% of interface requirements in the data centre are beyond 150m. Without the advent of extended reach 40GbE modules, data centre operators would need to deploy single mode fibre and a 40GBASE-LR4 module, it says. And while that can be fitted inside a QSFP, its power consumption is up to 3.5W, compared to the 1.5W of the QSFP+ eSR4. "The cost of the LR4 is also increased by at least a factor of three," says Tan.

Avago says that some 95% of all fibre in the data centre is multimode fibre. As for OM3 and OM4 deployments the ratio is 80% to 20%, respectively.