31 ms·
A big part of implementing the shuffle correctly at a large scale is having a good distributed filesystem upon which to build.
by timr 12y ago
A big part of implementing the shuffle correctly at a large scale is having a good distributed filesystem upon which to build.
- dsymonds 12y agoSure. I didn't mean to imply a good distributed FS was unnecessary or easy. Just saying that MR isn't easy.