Presentation is loading. Please wait.

Presentation is loading. Please wait.

ROOT Application Area Internal Review September 2006.

Similar presentations


Presentation on theme: "ROOT Application Area Internal Review September 2006."— Presentation transcript:

1 ROOT Application Area Internal Review September 2006

2 Progress Acknowledge the work done and progress achieved on the ROOT/Seal merger Good release schedule Good documentation –Need to integrate some new package doc (eg. Minuit2, RooFit) Very responsive to support requests

3 Packaging Compartimentalization of Run and Link time dependencies is there… –Do not link against GUI when not needed Split ROOT in packages that can be build and released separately –example: CORE, I/O,MATH, GRAPHICS, GEOMETRY/EV DISPLAY, PROOF –Setup a standardized package by package build procedure –insures rapid deployment and easier/faster uptake of eg. maintenance,bug fixes, new features Surely will improves ‘TTU’ –ability to built minimal applications –split 'implementation libraries' / 'dictionaries' when possible

4 CINT/Reflex migration Still very much needed, can't be left 1/2 done; Appreciate the honest presentation of delay, good explanation of work involved Careful (if slow) process Looks like the planning is now on a more solid basis than at the time of the last review But very ambitious, few experts. Good Luck! Ensuring continued binary compatibility will hopefully (should!) ease the experiment's uptake of Reflex

5 More Core pyROOT: several projects using it but single experiment-funded developer –Long term maintenance issue, eg. migration to Reflex Plugin manager convergence: –not much happened -- but it is still important/relevant! Packaging, dependency management, future maintenance –Ensure experiment buy-in by setting schedule, design in the AF As recommended during last review

6 The Technical Choices (iii) plugin management: substantially different approach: Factory (SEAL) vs. Interpreter (ROOT) carefully evaluate the impact on existing experiment schemes e.g. Gaudi component-model Esp. when visible to end-users From last year’s review

7 Math Much progress, impressed with the new fitting package enhancement, SMatrix, MVA, SPlot,… Future Plans as presented look reasonable

8 Miscellaneous Please keep names explicit/evocative –TRandom2  TRandomTausWorth –TRandom3  TRandomMersenneTwister Identify & reduce duplications, declare deprecated –Remaining SEAL code base, legacy ROOT classes –Needs of course close collaboration with experiments / users : Coordinate through AF Make full use of current C++ standard –Eg. TString vs. std::string

9 I/O Caching I/O, tree merging/splitting, –encourage further work on optimization in this direction Symptoms of communication problems? –CMS: SMatrix persistence –Atlas: Virtual inheritance in EDM –Ensure comm at all levels Beware of duplication of functionality ROOT & POOL projects –ROOT-RDBC/TFileSQL vs CORAL/POOL-Ora? Requested Functionality: –Thread-safety vs. active use of Multi-threading Reduction of use of global state Review of object lifetime policies –ATLAS request for RTAG Schema Evolution Stability is and must be paramount

10 PROOF PROOF is being used by Alice, deployed on reasonably large testbed.. Currently the only parallel interactive analysis environment –it is important that its development continues but in direct collaboration with the LHC experiment(s). Convert PROOF into a ‘product’ –predictable schedule for releases, –documentation and instructions how to deploy it as well as –detailed description of the architecture and required services so that important aspects of security can be understood by the security experts as well as the site administrators… Actively seek explicit endorsement of additional experiments –concept interesting, however, it would be desirable to implement a prototype that would mesh well with e.g. Gaudi –product that sits well at level of multicore CPU, local cluster, but not clear how useful/efficient it is when looking at very large data sets; in that respect, complementary to Grid. work on PROOF has other potential spin-offs: asynchronous de/ compression, xrootd,... which can be leveraged in Grid environment. –but be careful not to introduce complexity in order to satisfy requirements unique to PROOF.

11 GUI/Graphics –Many improvements, mostly asked for by users but which ones? LHC expriments? –Good for ROOT in general esp. as analysis package/presenter – But should not interfere with the role as core for the I/O, and foundation of the experiment software.


Download ppt "ROOT Application Area Internal Review September 2006."

Similar presentations


Ads by Google