Still alive, projects

August 19, 2012

If you wondered where I’ve been the last year, I’ve been on Google+ and somehow didn’t feel like posting anything useful in the last couple of months.

However, I’ve got quite a few cool projects running:

  1. Got me a pair of Infiniband Adapters. Plan: Brew up a software which will receive data via IP on multiple IB-nodes and write the data to distributed shared memory via RDMA/IB. Then let one or more IB-nodes read from that (ring)-buffer and aggregate data so that it can be shoved into an RDBMS.
  2. Brew up an AWS image for easy BOINC-crunching while preserving the workunits on a headnode. Why? Because images on the AWS spot-market are cheap. But spot-machine do not retain any data. So I might be putting it into S3 along with some headnode directing which machine (machines id keeps changing!) can work on which workunits.
  3. Become more confident in the language Erlang. Got quite a few projects which could benefit from easy protocol prototyping with ASN.1 in Erlang. Stay tuned.
  4. Got me a Spartan-3E FPGA board. Not sure what to do with it, but it’s awesome :)
  5. Was playing around with GNU Radio and my RTL DVB-receiver recently.
  6. Doing some serious Openstreetmap work recently.
  7. Mrs. Janssen and be got fond of geocaching. Lot’s of outdoor stuff. Sweet!
  8. Been to this year’s Linuxbierwanderung. Was awesome!

I ain’t dead yet.


A Critique of RDMA

August 24, 2006

f77Patrick Geoffray from Myricom, the creators of Myrinet, posted a nice rant about RDMA and why it fails to scale well with MPI and direct socket-communication. He basically says that every time you need to register or synchronize memory in MPI, RDMA looses all it’s advantages since the CPU is heavily involved: Those operations are expensive and you add latency to the operation.

Read his fantastic rant – backed by statistics – at HPC-Wire.

Tech Tags: