A Critique of RDMA

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:

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

%d bloggers like this: