Object storage targets are an approach but has anybody tried it out and it worked?
Curious as to what you're referring to. I Googled "parallel m:n i/o". there were a number of papers found, but of those a couple might be helpful to you, although I neglected to mention "object storage": http://www.mcs.anl.gov/~thakur/papers/sc11-io.pdf http://www.sandia.gov/~kmorel/documents/SciDAC08-2.pdf
I've been retired for quite a few years, so can't really offer any help myself...
Thank you James,
the first paper I knew already.
the second one from sandia is new to me and might help to get more insight.
We are trying here to improve our I/O. Load balancing is good but there is at last one node responsible for writing out to disks which is a real bottleneck.
Thank You for helping even when you are retired.
- Badges
- Science topic
- Similar topics
- Computer Science and Engineering
- Software
- System Software
- Operating Systems