Browsing for deadlines, I found a potentially interesting special issue for JSAC http://www.jsac.ucsd.edu/Calls/switchingandroutingcfp.pdf which proposes the following timetable

• Submission:           January 1, 2013
  First review:         May 1, 2013             (4 months)
  Final decision:       July 1, 2013            (2 months)
  Final to publisher:   September 1, 2013       (2 months)
  Publication:          First Quarter 2014      (4-7 months)
  

The entire process takes 12-15 months from submission to publication. In my opinion this is way too long for a process where everything can (and should) be planned in advance.
To be fair, other JSAC issues have slightly better times, some even down to 8-11 months. But still there is no justification for a 6-month review period (I am talking about review because this takes out the allegedly long times for typesetting, proofreading and printing).
A long turnaround time is harming the journals in the first place, as they will publish obsolete results and scare away students (who need short-term feedback) and the most productive researchers who have better venues (conferences) to publish their results and move on with new work.
Without even sacrificing quality, the review cycle could be trivially reduced to 1 month, using strategies that many other professional organization adopt:

• contact reviewers in advance and only enroll those who commit to complete a first round of reviews within 1-2 weeks from the assignment, and another smaller round in the upcoming week. Pay them a nominal fee, if needed: after all, reviewing is a professional activity, scientific publishing is often a for-profit activity, and if costs (which are modest) need to be cut, one could start with removing physical printing and moving to fully online);
• same as many conferences do, ask authors to submit title and abstract one week in advance, so allocation of papers to reviewers can be done in parallel with the final submission

This gives 2 weeks from submission to completed reviews, and another two to run an online discussion and second round of reviews. Also this means that feedback to the authors arrives when they still remember something about the work, and not 1-2 years later, and possibly there is still time to run more experiments and improve the paper.
Of course a key element is the professional handling of the process, which means that the editor in the first place needs to closely monitor the work of the reviewers, and have sufficient redundancy and/or backup plans in case someone is late.

We have just finished a first prototype of VALE, a high speed Virtual Local Ethernet that can be used to interconnect Qemu instances but also network applications that talk the netmap protocol (or libpcap clients). You can find code and papers on the VALE homepage

With Marko Zec and Miljenko Mikuc we have completed a paper reporting on a very fast scheme called DXR for IPv4 lookups, see Towards a billion routing lookups per second in software We still hope that there is a use for this work, even though many believe that IPv4-related research is archeology.

While patiently waiting that the IEEE/ACM ToN completes the review of our QFQ paper (submitted in Nov.2010), with Paolo Valente we studied how the internal FIFO queues in the NIC (and possibly in the device driver) affects the scheduler. You can find the analysis in this recent paper, Analysis of fair queueing schedulers in Real Systems