Out of the office

Now that AdvancedTCA is enjoying wide scale deployment in the Central Office and is grabbing market share from older, less capable, and mostly proprietary architectures, some unexpected things are happening. AdvancedTCA was designed from the get-go for telecom Central Office environments. NEBS requirements shaped many design decisions, including minus 48 VDC power, 55 degree Celsius operation, comprehensive system management, field replaceable modules that can be swapped while the system is operating, and more.

An interesting thing is now happening – something quite unexpected at least by this columnist. Some of the architectural features, primary among them the high availability capability of AdvancedTCA, are attracting the attention of a wide range of developers and integrators in very diverse fields. A few months ago I wrote of the interest on the part of the “big science” community, including accelerator physicists and fusion researchers. That community has now formed three different PICMG technical subcommittees working to solve their unique problems. Committee representatives from major labs worldwide are participating. AdvancedTCA is already flying on the next generation submarine hunter aircraft, the P-8 Poseidon. And it is also being considered as a replacement for tens of thousands of servers used shipboard by the U.S. Navy, as Brian Carr of Emerson Network Power details in his article in this issue.

There is another application that AdvancedTCA is well suited for if a few things are changed, and that is as a replacement for racks and racks of 1U servers typically used by data centers. That community is interested in the high density and serviceability of bladed architectures like AdvancedTCA. These folks have been turned off in some cases by the relatively high cost and proprietary nature of the bladed offerings available today. An open architecture like AdvancedTCA has the potential to reduce costs if volumes are high enough and competition is enabled. So PICMG has started up a technical subcommittee, known as AdvancedTCA Extensions, to create suitable versions of AdvancedTCA for use in data centers and other non-Central Office environments.

The 200 watts per slot power limit normally associated with AdvancedTCA in a Central Office environment simply isn’t enough, however, for the high power processors and the large amounts of memory required for data center applications. AdvancedTCA is capable of significantly higher power levels if a few things are changed and a few requirements are relaxed. Data centers typically operate at a maximum ambient temperature of 35 degrees C, not 55 degrees C with a maximum temperature rise inside the chassis of 10 degrees C as required by Central Office NEBS compliant designs. Using cooler air and allowing a higher internal temperature rise in an AdvancedTCA chassis raises the useable power levels significantly. Another way to increase power levels, provide more performance, and reduce costs is to standardize a few double-wide blade designs. If the normal 1.2-inch board pitch of AdvancedTCA is doubled to 2.4 inches and is made to occupy two slots, twice as much airflow per blade is available. Higher power processors can be used. Also taller – and much cheaper – DIMM memory manufactured in very high volumes for conventional PCs and servers helps reduce costs. The AdvancedTCA Extensions subcommittee is working to standardize at least two double-wide implementations, and each is optimized for slightly different applications spaces. The power goal for these new modules is 800 watts, and it looks like that can be achieved. The power connector used in AdvancedTCA has a large amount of margin built into it, and it appears that new backplanes with larger traces and thicker copper will get to that power level. Tests by several of the power connector manufacturers are already underway.

Two articles in this issue relate directly to this effort. Brian Carr goes into some detail about the various non-Central Office applications for AdvancedTCA that his company is seeing. David Wright, Advanced Platforms, continues his excellent tutorial series on cooling AdvancedTCA boards and as usual, his article is packed with careful engineering backed by meticulous references.

Applications like IPTV are driving the need to increase the already high backplane bandwidth of AdvancedTCA. Currently a full mesh 16 slot system has, at least in theory, a maximum data transfer bandwidth of 2.5 terabits per second. That’s a lot, but never enough for some, it seems. PICMG is working diligently on adapting the new 10 Gig KR standards to quadruple that. That will require very careful design of both boards and backplanes. Tim Lemke from Z-Plane is an expert in high-speed signaling and he and co-author Justin Moll, Elma, share with us some new technology invented to simplify the design of these higher speed backplanes and lower their cost using perpendicular “riser” boards that carry the 10-gigabit-per-pair signals.

This journal typically covers a wide range of topics, and this issue is no exception. My former colleague Gary Drossel, Western Digital Technologies, gives us some insight in the uses of solid state disks and provides useful lifetime calculations in comparison with conventional rotating media.

Hot swappable architectures such as AdvancedTCA and Advanced Mezzanine Cards need a way for the system to know when a board is being inserted or is about to be removed. This is often accomplished by using specialized microswitches. David Webber from C&K explains to us what is important in selecting these devices and why.

Enjoy.