DAIS Telcon 25/03/04 ==================== Chair: Norman Paton Minutes: Mario Antonioletti Present: Mario Antonioletti, EPCC Amy Krause, EPCC Norman Paton, University of Manchester Susan Malaika, IBM Brian Collins, IBM Simon Laws, IBM Vijay Dialany, Southampton University Neil Chue Hong, EPCC Allen Luniewski, IBM Tod Guay, DMTF/Oracle Thomas Soddemann, RZG Steve Langella, Ohio State Shannon Hastings, Ohio State Tashin Kurc, Ohio State Savas Parastatidis, Newcastle University Joel Saltz, Ohio State Agenda 1. Minutes/Matters Arising from 4th March Telcon. 2. Matters arising from GGF11. - Charter. - File access. - Object database access. - Data in OGSA Data Architecture Document. 3. Updates on document planning moving towards GGF11. - Core Specification. - Realisations. - Request composition. - Mapping DAIS to WS-* proposals. 4. AOB Actions Resulting from the Meeting: [Thomas & Norman] interact to come up with a page description to scope out a OO database characterisation. [Brian/Neil/Ally] To interact with each other with regards to coming up with a file access BOF for GGF11. [Norman] inform the data area chairs to inform and discuss on how to proceed with regards to File/OO database access. [Norman] set up a telcon to discuss a telcon for object database access... establish a time and then publicise for people who are interested. ------ Norman welcomes participants and explains that a bunch of people are currently in Manchester discussing the specs. Minutes and matters arising from the 4/3. No actions from this meeting. Some actions from GGF Thomas: spoke to people at signsoft regarding objects. They are extremely interested in this group. Norman: also had some contact from Objectivity. Need to see what should be put on the charter but also need to set some process for the people that expressed an interest and can collaborate... do you have thoughts? Thomas: have contact with Ammar ... can start working ... not sure about the charter .... need some help... Norman: need time scales and deliverables ... need to also ensure that the people that are participating in this area see things in the same way ... should set up a conference call for people that are interested in this and use this a mechanism to get going. People may also want to start interacting ... Simon: as part of this process we should gauge the time scale of that activity ... we need to explain what forms the other specs will take. Norman: anything else Thomas? Thomas: maybe we should also discuss what we should do with distributed objects on the Grid ... Norman: this is a hot potato ... distributed objects is a term one has to handle with care in this settings ... service based access to distributed object and this going on for ever ... Thomas: Java, RMI ... if we could come up with something so that updates could be done ... that would be quite a good idea... Norman: looking at objects in a Grid setting ... this would be outside the scope of DAIS ... the ethos in DAIS is to try to provide service orientated access to data resource with the query language coming from that data resource and also that we can use consistent terms ... and that is what DAIS seeks to achieve. A more general framework would challenge our charter... Thomas: we should have the more general pattern in mind when discussing the data access to OO databases... Norman: we need to have a little more dialogue on this... we should also you and I have should have some interaction off-line to see what a database realisation for objects should look like within DAIS .... Norman: file access.... Neil: I am here as a proxy for Ally Hume ... I work on ogsa-dai and we've been involved in this area before through Martin Westhead ... we've done some work on this in ogsa-dai too. We do not see much of a difference between file access and the other types of access ... we've been doing some work on ogsa-dai and would like to standardise this ... Norman: within GGF there is GFS and there was also a BOF run by Peter Kuntz and someone else .... someone from LBL ... Simon: storage management systems ... Norman: In terms of file access are you thinking of things that look inside a file or don't Neil: kind of both ... could query within files and across files Vijay: what is the data model? Neil: considering fairly simple models ... table... Vijay: relational ... Norman: does this mean you are thinking of cvs (comma separated values) Neil: that would be one example ... and also xml files ... there is an overlap the xml realisation and more loosely based xml files in ... Joel: that is interesting ... we have a project called storm in conjunction with the mobius project ... have distributed collection of files in storage systems and using the metadata can target SQL queries .... map it to collection of files ... similar virtualisation ... it is a motivation in distributing the software through the national middleware ... we would like this to be canonical ... the general concepts that you should be able to query files on the Grid regardless of the type of data that you are trying to access... Allen: in both cases trying to present a relational view of files ... why not use the relational realisation of DAIS? Joel: one would not have to use the relational ... ... Neil: you can consider other formats that skew off the relational view ... you want to try to capture the general case ... there are some general files that can be applied a whole set of structured files .... Allen: I think of DAIS as being about how access data, ie. the query language, and not how the data is stored. So, we can relational access to data independent of whether it is stored in an RDBMS, an XML repository or text data in a file. Thus in the DAIS context I think of file as being an access method that looks like file access (e.g., Open, ready bytes, close). Joel: so the mention of image in the medical domain ... groups interested in maintaining large number of images on-line and processing these images ... these will be in different formats and stored in different ways ... the issues in what should/should not be standardised - what can you say of files in terms of attributes and formats and ... how you can interrogate these data collections ... which should be standardise is subject to some debate ... but some of it should be ... Norman: thanks for that ... we've shied away from composition ... dais specs concentrate with access but have not gone on to the relationship between their relationships in a federation ... Susan can you comment (as an author of the gap analysis)? Susan: we did not do a very detailed investigation of files or the types of data that you may want to model in GGF .. there was another similar topic which is the metadata that is used to model the files ... have some data in files and metadata in other files ... did not hear that mentioned ... and that came out strongly in the files... Shannon: SQL based access to files seems to take a step in that direction... Susan: was not saying that you want to access files in SQL .... the description of files is kept in relational databases and that needs a special treatment in itself ... Allen: I kind of understand what they are trying to get ... an access style to files that encompasses various languages ... Susan: that would be interesting to see how we would treat that ... ... Shannon: is there a way to describe file in general and if there is a way to access a file that does not require a relational way... ... Joel: ... one way to do is to have metadata that describes parts of files that allows you to access parts of the files ... Susan: is that not what the DFDL group does? ... Shannon: DFDL does that but what is missing the access pattern ... because of DFDL we could do this ... Susan: ... the other area that always comes up is streams.... Shannon: there would be similar access patterns Susan: there a few types of info that we have not dealt with ... ... Norman: it's quite clear what happens in this area ... either nothing or a BOF... the BOF would identify precise topics on which people would be interested in collaborating and the people would be interested in collaborating ... need to identify the people that would be interested in setting up a BOF for GGF11 Shannon: are you saying that file access does not belong in DAIS? Norman: the data area chairs have told us that if file access is going to be supported then a BOF needs to be done and the framework to set up on the grounds that not everyone that has an interest in files would be in DAIS ... the incorporation of files in DAIS would be an option that could be considered by the BOF but we have been told that there should be a BOF ... are there people that are interested to do this at GGF11? Brian: following from what Susan said ... I have a lot of interest whether the metadata is in relational and the files as BLOBS ... I would be interested in participating in a BOF for GGF11... with the caveat not on my own ... Shannon: Martin would want to participate ... Neil: it was his intention to take the work that he had been doing to a sensible point ... keen to be involved ... Martin will be leaving EPCC but he is still keen to carry work on some of the GGF work ... he will be handing over this ... currently Ally Hume and I would be keen to do this ... Norman: Ohio ... Joel: we need to discuss this off line ... quite possibly we may get involved ... Norman: Allen would you like to report on the OGSA Data Architecture document and the ogsa data service part of that ... anything to report on that? Allen: there was a telcon involving that group ... getting organised ... a collection of people ... work should be completed no later than GGF11.... Norman: any impact on DAIS? Allen: nothing immediate other than trying to give some input ... the straw man proposal being put together is not going to be useful outside of this group ... trying to make sure that we all speak the same language Simon: who is in this group? (Allen goes off to check ...) Savas: why is this not being done in the DAIS WG? Norman: the scope of that work is bigger than the scope of DAIS... would include replication and data movement ... an umbrella activity.. Allen: Peter Clarke, Andrew Grimshaw, Dave Berry, Ian Foster, Arun Jagatheesan, Osamu Tatebe, Malcolm Atkinson, Inderpal Nerang, Allen Luniewski Norman: as a group we stay put and wait until something is circulated round for comment.... ... Norman: mapping of DAIS to WS-* Simon: the process is on-going ... had a call last week ... get ready to distribute to GGF11 Norman: there is not going to be another workshop at the next GGF but will have a DAIS session ... Savas will be there.... there should be a tidied up version of the doc that was discussed at GGF 10 within a couple of weeks ... ... Norman: DMTF ... the DAIS group needs to describe a lot of metadata relating to data services .. the DMTF is interested in describing objects in a distributed world ... DAIS have had some discussion with the DMTF people ... trying to ensure that DAIS metadata will be more pertinent than DAIS .... had a combined session at GGF Susan: had some comments from the scenarios doc from Tod .. the CIM people need to get a group of people to work on this with the DMTF... Tod: updated version looks a lot better ... Susan: you looked at an early version Tod: will mark up the current version ... Susan: that would be good Tod: in DMTF we concentrated on the content in rfc1697 (?) ... (other work described) ... additional areas from looking at your doc there are other areas we have not covered ... do not have that many people involved in our group ... they also come and go ... different groups of people have been involved ... we are starting on cluster information ... database clusters ... Susan: do you think you could get stuff in 2.9 for schemas ... Tod: 2.9 has to be submitted for next Tuesday.... Susan: are most of the people that are active database people or management ... Tod: I am a management person that also do database ... systems management .... Susan: we will not make 2.9 preliminary which we'll not make 2.9 preliminary.. Tod: when 2.9 closes there is some review by companies ... do not want to put new models after it has going preliminary ... Susan: should Larry arrange a call? Tod: have a call next Monday ... could discuss how to work together .. can send Larry an email message ... Norman: that would be a good thing ... have several ... Tod: that is a DMTF meeting... need to flag if you are interested to work with us ... Susan: you need to get Larry to put it on his mailing list ... Norman: We are not in a rush ... there is still a lot of work still to do... we should be able to accommodate this in DAIS ... ... Norman: AOB? *silence* Norman: we have a call in a week ... will try to get an agenda a bit further in advance ... thanks for coming ..