Choose timezone
Your profile timezone:
ROOT I/O Meeting
Attendees: Philippe, Oksana, Jim, Jakob, Guilherme
Oksana: Interesting CHEP conference. Mystery (lossy) compression algorithm with good/amazing result. I felt that the poster could use additional exposure/presentation. More than 100 other poster some very cool (eg an on tablet 3D medical imaging). Module talk was very well attended and received.
Oksana: you can preview the slide version of the compression algorithm poster at https://docs.google.com/presentation/d/1tXjnS-o1Vf6F6OrJXpFNNY0x5Cu0KgFYOpjz_0A7Xh8/edit?usp=sharing
Jakob: Nice trip including driving from Adelaide to Melbourne. Wild life was very interesting, very unusual.
Jakob: RNtuple talk was well received, lots of good comments, was included in the summary slides. Several common questions. Gordon W. and al still pressing the point of making independent library. Andrea Rizzi talked about compression, including lossy. Indexed friend tree was another important request.
Jakob: Andrea Rizzi plenary on Data Analysis for Run3. Markus Schultz talk about data management, how data is replicated and used, 38 slides for 12 minutes. Talk on computing cost model. Atlas OpenData (ntuple version). Strange talk for CLASS12, the HIPO data format (for online). Row-wise even storage with some indexing. Claim a factor 3 to 5 better than ROOT but the ROOT version of their example was significantly sub-optimal. Benchmark was clearly not done carefully (eg. HIPO is faster to read 2 variable than 1 !?!?)
Axel: I will contact: Volker Burkert from CLASS12 to see if we can go over the tradeoff between their own solution and ROOT.
Jim: ktai a tool to reverse-engineer data format. Part of a different culture in nuclear physics, more fragmented.
Jim: I shown some RNtuple result to some people and those were excited. I had a thought was listening to talks, after being done with AwkwardArray v2, what about I write down some documentation about the ROOT file format; this was a bit inspired by Jupyter notebook; writing down an information document mixing text description and code fragments. (Would start sometime in the Spring).
Guilherme: G4 I will try to flip the loop. Instead of doing a full particle, instead doing each time for all particles in flight (and sorting particle per type).
Axel: module and recursive execution.
Philippe: I took a look at TTreeCacheUnzip .. It needs a bit of sprucing up to become (really) useful (including adding parallel prefetching)
Oksana: Going to Kiev for a couple of week to receive my degree. I will make a presentation about ZSTD and talk about TTree and RDataFrame at the Software Carpentry.
Jakob: back to RNtuple, in particular Header and Footer (docs and improvements). RNtuple workshop to talk with the data center maintainer to discuss what we can design to improve the efficient.
Jakob: Mid-February dateline for CHEP papers.
There are minutes attached to this event.
Show them.