ROOT I/O Meeting

Friday 31 Mar 2017, 15:55 → 17:20 Europe/Zurich

VIDYO

Attendees: Viktor, Zhe, Philippe, Axel, Jim

Zhe presented the result from LZ4.  Compression and Decompression is significantly better than Zlib (and even more significantly better than LZMA).  However compression ratio is smaller.  From slightly less than 10% for one of the CMS yardstick files to 50% for a synthetic example.   The compression ratio different is less for smaller baskets than for larger baskets.  (And the yardstick file has a large fraction of small baskets).

Philippe/Axel: In order to introduce LZ4 (or any new compression algorithm) we need to make sure that there is a clear use case (to avoid unnecessarily increasing our support load).   In this case, it unclear whether the extra storage need will be a show-stopper for experiment to adopt it despite the CPU gains.  It is also unclear whether individual physicist will know to turn it on (as it will benefit them the most directly) and turning it on by default would break forward compatibility (unless we back port lz4 to holder releases …).    
Let’s get feedback from the experiment.   Zhe can you present at the April 12 ROOT planning meeting?

Zhe: Yes, I will be happy to.

Axel: Could you enhance the presentation by giving a plot showing the algorithm vs ratio vs speed.  Something like http://sd-1.archive-host.com/membres/up/182754578/Enwik8-bench.png or http://dbahire.com/wp-content/uploads/2014/08/compression2.png

Zhe: Yes, I think so. 

16:00 → 16:20 Final Update on LZ4

Speaker: Zhe Zhang

lz4presentation.pdf

16:20 → 17:00 Round Table