Comments (3)
  • Add a Comment
  • Edit
  • More Actions v
  • Quarantine this Entry

1 Ajay Dholakia commented Permalink

This is very good news ...

It would be good to understand the factors contributing to this gain ... Cores, dataset size, scheduling, etc. are all different. 10 TB vs 1 TB. 18 12-core data nodes (216 total cores) vs 9 16-core (144 total cores), without vs with optimized scheduling ... Is the GB/min/core number an appropriate metric for comparison?
Can you share some more information?

2 Anirban C commented Permalink

Ajay, thanks. We'll be working with a third party to formally write up the results ... that write up will contain more of the test environment details you're looking for. Stay tuned!

3 Jake Kitchener commented Permalink

I'm eager to learn more about the details of your configuration. I'm assuming that more details will be included in the formal write up?

- Disks Used?
- Networking?
- Complete software stack? You mentioned Symphony, were you using Symphony MapReduce here?