Micro QSFP module to boost equipment port densities  
Wednesday, September 9, 2015 at 7:04AM
Roy Rubenstein in Micro QSFP, Nathan Tracy, QSFP, SFP, optical transceivers, thermal capacity, μQSFP

Twelve companies are developing a compact Quad Small-Form-Factor Pluggable (QSFP) module. Dubbed the Micro QSFP (μQSFP), the multi-source agreement (MSA) will improve by a third the port count on a platform's face plate compared to the current QSFP.

 

Nathan Tracy

The μQSFP will support both copper and optical cabling, and will have an improved thermal performance, benefitting interfaces and platforms.

“There is always a quest for greater port density or aggregate bandwidth,” says Nathan Tracy, technologist at TE Connectivity and chair of the μQSFP MSA.

The challenge for the module makers is to provide denser form factors to increase overall system traffic. “As we go to higher densities, we are also increasing the thermal load,” says Tracy. “And so now it is a mechanical and a thermal [design] problem, and both need to be solved jointly.”

The thermal load is increased since the μQSFP supports interfaces that consume up to 3.5 W - like the QSFP - while having the width of the smaller SFP rated at 1.5 W. 

“We are limited in the directions we can pull the heat out,” says Tracy. “If we are going to enable a higher density form factor that has the same width as an SFP but it is going to have the functionality of a QSFP, now we have a problem.”

This requires the MSA engineers to develop new ways to rid the μQSFP of its heat.

 

If we are going to enable a higher density form factor that has the same width as an SFP but it is going to have the functionality of a QSFP, now we have a problem

 

Heat transfer and other challenges

The volume and surface area of a module determine the overall thermal capacity or thermal density. The module can be modelled as an electrical circuit, with heat flow equivalent to current, while each interface has a thermal resistance.

There are three interfaces - thermal resistances - associated with a module: between the heat source and the module case, the case and the heat sink, and the heat sink and ambient air. These three thermal resistances are in series and the goal is to reduce them to ensure greater heat flow.

The module’s circuitry generates heat and the interface between the circuitry and the module’s case is one of the thermal resistances. “You are going to have a heat source in the module and no matter what you do, there is going to be some thermal resistance from that source to the module housing,” says Tracy.

 

You have to get good signal integrity through that electrical interface because we are working at 25 gigabit-per-second (Gbps) data rates today and we know 50 Gbps data rates are coming

 

The second thermal resistance - one that the µQSFP eliminates - is between the module housing and the heat sink. Sliding a module into its cage puts it into contact with the heat sink. But the contact between the two surfaces is imperfect, making heat extraction harder.  Building the heat sink into the μQSFP module avoids using the sliding design. 

The remaining thermal resistance is between the heat sink and the cooling air blown through the equipment. This thermal resistance between the heat sink's metal fin structure and the air flow exists however good the heat sink design, says Tracy. 

Other design challenges include achieving signal integrity when cramming the four electrical lanes across the µQSFP’s smaller width, especially when its support 25 Gbit/s lanes and likely 50 Gbit/s in future, says Tracy.

And the module's optical interface must also support duplex LC and MPO connectors to interoperate with existing cabling.

“It is all a balancing act,” says Tracy.  

 

Applications

The μQSFP is aimed at platforms such as 4.8 and 6.4 Tbps capacity switches. The QSFP is used for current 3.2 Tbps platforms but greater port densities will be needed for these next-generation platforms. The size of the μQSFP means 48 ports will fit in the space 36 QSFPs currently occupy, while 72 μQSFPs will fit on a line card if three rows are used. 

The μQSFP may also find use outside the data centre for longer, 100 km reaches. “Today you can buy SFP modules that go 100 km,” says Tracy. “With this form factor, we are creating the capability to go up to four lanes in the same width as an SFP and, at the same time, we are improving the thermal performance significantly over what an SFP can do.”

The Micro QSFP group is not saying when the µQSFP MSA will be done. But Tracy believes the μQSFP would be in demand were it available now. Its attraction is not just the greater port density, but how the µQSFP would aid systems engineers in tackling their thermal design challenges.

The pluggable form factor will allow air to flow from the face plate and through the module to where ICs and other circuitry reside. Moreover, since 32 μQSFP ports will take up less face-plate area than 32 QSFPs, perforations could be added, further improving airflow.

“If you look at the QSFP or SFP, it does not allow airflow through the cage from the front [plate] to the back,” says Tracy.  

The μQSFP MSA founding members are Avago Technologies, Broadcom, Brocade, Cisco, Dell, Huawei, Intel, Lumentum (formerly JDSU), Juniper Networks, Microsoft, Molex, and TE Connectivity. 

 

Article originally appeared on Gazettabyte (https://www.gazettabyte.com/).
See website for complete article licensing information.