Meeting Notes for 13 February 2008

Present: Eric, Liz, Mihael, Tom L., Joao;
Absemt: Dan K, Mike W., Randy., Kris, Joao, Tom J, Bob

Joao's Report on ATLAS e-Lab and Amelia

Amelia stands for ATLAS Multimedia Educational Lab for Interactive Analysis.

Phase 1: Basic Framework of Amelia

Amelia shifted to multi-platform approach and had to redesign from scratch. It works on Mac, Linux and Windows and has a new user interface. It is much easier to use. They will be doing their last alpha at the end of the month. The analysis on the local machine running Amelia involves pattern recognition looking at event displays. Students learn what makes a Higgs event a Higgs, etc. He built his own GUI system based on IRRLICHT graphics engine. (See http://en.wikipedia.org/wiki/Irrlicht_Engine)

Phase 2: Grid-enabling of Amelia - data-gathering

Next they will work on the data-gathering where they will load events by file. Events have to be in a certain folder. Collect statistics on these events. Use the computing power of the grid to select the events from CERN and do the statistical analysis.

Two modes of gathering events from the pool of events from ATLAS (either at CERN or at one of the Tier 1 or 2 sites)

  • Basic mode: where download is automatic to the users' computer
  • Advanced mode where users can make choices and get the events. Tweak the query to the grid for the type of events you want. Users will be able to save a file with the date of their analysis and available to others so they can will be able to reproduce it. (Liz: in the discussion there was some ambiguity between whether the term "analysis" referred to the activity of selecting events or the activity on the local machine to study the event. We need to find more precise vocabulary for these activities, and for the data repositories used).

Other features:
  • Guided Tours – 3 K files where you can have an automated way to see what someone else does while looking at the event. Guided tour recorder. Mouse click and track selection. Anyone can load that.

  • Setting up templates where the program automatically tries to find events with a higher probability of being interesting.

Firefox engine (Gecko) inside Amelia. Tutorials can be located on a central server. Students can actually access the web from Amelia and do not need a web client.

Liz: The e-Lab philosophy is that this part should be implemented so that students can use standard web clients and not be dependent on an application on the local machine.

Go to the pool, gathering data, keep track of the result of the analysis. These will be flagged. This will bias the distribution so other students will get them. Hall of fame of interesting events. Anyone will be able to download these. We can make guided tours for these interesting events.

Data comes from a database connection to a database with a huge pool of data at CERN (Tier 0) or BNL (Tier 1), or from ATLAS tier 2 sites (there are 5). User asks for a certain event through the e-Lab. The CPU intensive software running on the grid needs to filter among all these billions of events for a set that is appropriate for the user. We need to index all these event and put them in a database. Then build a smaller database with information and events for students that we distribute. Want to put online list of events that had more positive analysis. There will be a Hall of Fame of events.

Publishing: make a poster using the e-Lab. Upload capability needed. They could also publish a "guided tour".

Testing: According to Joao, no one has put up any objection to downloading the events and the application to the local machine. You could run it from a CD or just from the desktop. Later in talking with Kris, it seems like Joao has tested this with a high-school student and undergraduates who work with them at LBNL, so it is not clear whether actual schools have been involved yet and it may be premature to say that schools will be able to have software on their school computers. Thus they may need for the next phase for many schools.

Joao responded to this point in a subsequent email: "We have NOT test-driven the application with any school yet. What we did was: a) involve a few undergrad students and a high-school student in the development and assessment process; b) show an early alpha version of the application to selected groups of education specialists and request their input about its usability, features and interface; c) informally ask the same selected groups about eventual obstacles they could see about the usage of such an application in schools. From this sample, no objections were found about the eventual usage of this kind of software in their computers, specially since the way the new design is being implemented does not require an actual installation and can even be run from a non-disk drive (such as a CD-ROM or a flash drive)"

Eric pointed out that the students will be acting as a "Mechanical Turk" -- named after a machine built in the 18th century which purported to play chess, but actually contained a person hidden inside. A modern example is the Amazon Mechanical Turk, where people can be paid to do tasks a computer cannot do well. One task on the Amazon site was to search through satellite photos from Google Earth for evidence of downed aviator Steve Fossett (details). Eric has set up a Mechanical Turk using reCAPTCHAs, where humans decipher words which cannot be interpreted by computers as part of a project to digitize old books. See http://pirates.spy-hill.net/top_turks.php

This lead to a brief discussion of the old days of particle physics when human "scanners" were hired to view pictures taken from bubble chambers. It might be useful to somehow tie in that history to the more modern Amelia.

Phase 3: Amelia-Lite – completely web-based.

Joao and his team are thinking about doing Amelia-lite that will be java-enabled. It is really hard to do the 3-D stuff in Java. Stage 3 will be less interesting (Java-enabled). The geometry will be have to be less complicated.

Amelia and ROOT: Want to avoid the client having to install ROOT so Joao has made his own software.

Eric: Getting ROOT on the desktop is desirable. But it cannot be required, since we cannot assume we can install anything on the desktop, so it would only be in addition to a web based tool. He also suggested that it might be possible to use ROOT via the web based tool he's developed for LIGO, but with ATLAS-specific ROOT code.

Joao reiterated the basic difference between the way CMS and ATLAS e-Labs are designed:
  • CMS - Statistic analysis of huge datasets.
  • ATLAS/Amelia - Per event basis – Amelia; Filter the data according to a query.
He pointed out that this was intentional, to make the two complimentary.

Tom L. raised the question of identification reliability. How certain are we of individual student’s results? It will involve the agreement of a number of students. You could watch the dynamic of this. Talking about this issue with students would be educational.

Eric suggested that Joao and/or Mihael might want to talk to Rob Gardner at the CI, since he leads weekly telecons for US ATLAS grid efforts. ATLAS does not use SWIFT, but Mihael can probably quickly sort out what needs to be done and how to do it with SWIFT. It would be useful to be able to "borrow" as much as possible from existing ATLAS grid code, rather than having to create it ourselves. Mihael says he knows people in Gardner's group, so it will be easy to make contact with them.

Action Item: Joao is going to make a diagram so that everyone can understand the different parts of this system. This should help Mihael understand how this fits in the e-Lab structure and help Joao understand how he could use the analysis tools Mihael provides to execute jobs on the grid. We will put this diagram on the web and provide a place for Joao add documentation.

Action Item: Eric will send an email to Joao about where he has documented getting accounts on the CI Wiki (ie ForNewMembers). (DONE)

Tom wants to get Dan K. to join in the discussion of this next week. He thinks it would be useful for him and Dan to visit Joao at Berkeley.

Announcements:

March and half of April - Mihael will be gone.

Eric may not be available in the next week or two, maybe a bit longer.

-- Main.LizQuigg - 14 Feb 2008
-- Main.EricMyers - 14 Feb 2008
Topic revision: r13 - 2008-02-14, LizQuigg
This site is powered by FoswikiCopyright © by the contributing authors. All material on this collaboration platform is the property of the contributing authors.
Ideas, requests, problems regarding Foswiki? Send feedback