- Status of JIRA and Bitbucket
- Latest survey results (closed on October 23, 2019; 16 anwers)
- Summary of last meeting
- How to set goals for Board: address one but the most important item from list of weaknesses (we identified comprehensive use cases and code examples from Mark's list).
You should have the following invites (all times Central):
- 11/5/18, Monday, 11:30 am - 12:30 pm (today)
- 12/6/18, Thursday, 3:00 pm - 4:00 pm
- 1/7/19, Monday, 11:30 am - 12:30 pm
- 2/7/19, Thursday, 3:00 pm - 4:00 pm
- 3/4/19, Monday, 11:30 am - 12:30 pm
- 4/4/19, Thursday, 3:00 pm - 4:00 pm
- 5/6/19, Monday, 11:30 am - 12:30 pm
- 6/6/19, Thursday, 3:00 pm - 4:00 pm
- 7/1/19, Monday, 11:30 am - 12:30 pm
- 8/1/19, Thursday, 3:00 pm - 4:00 pm
- 9/2/19, Monday, 11:30 am - 12:30 pm
- 10/3/19, Thursday, 3:00 pm - 4:00 pm
- 11/4/19, Monday, 11:30 am - 12:30 pm
- 12/5/19, Thursday, 3:00 pm - 4:00 pm
|40 min||Continue review of strength and weaknesses||All|
Discussion summary (for meeting details check video and transcript):
- We agreed that response to the strength and weaknesses survey was not good. There was a suggestion to ask other Board members if they have experience of reaching out to a wider audience.
- We acknowledge that scheduling challenges for the Board. Mark suggested to that full Board get's together at least twice a year. Elena will follow up.
- Charter discussion - we decided to sent Charter and vote via email.
- Continuation of strength and weaknesses:
- HDF5 cannot be tuned for specific patterns like log structured writes (for example, bunch of elements that are not contiguous in the data set address space, for example but in a log structured style you would allow those to be written continuously into the file).
- HDF5 and Object Store
- There is on-going effort to allow HDF5 applications to store data in RADOS or DAOS OS. The HDF Group will be release VOL architecture in March 2019.
- Suggestion is to come up with a set of use cases that are large-scale use cases which may be separate from learning exercise and figure out what performance people are actually getting and then let's see if we can come up with a handful of necessary improvements and tactics one of which might be log structure that would bring HDF5 up to par with these other libraries (ADIOS).
- Discussion of different motifs/examples users can chose from to avoid misuse of HDF5 library; Mark shared his experience working with the engineering group to improve HDF5 performance.
- Discussion of why THG need to address lack of architectural documentation (weakness). There are several refactoring efforts are going on now; it is easy to introduce new issues when one doesn't know what is being refactored and its affect on other parts of the library. We need to do some sanity checks and ask ourselves if what we are doing now will make things harder or easier in the future.
- We will continue discussion of strength and weaknesses with the goal to come up with the roadmap.