Time/Place
- Time: 3:00pm Eastern Standard Time US (UTC-5)
- Dial-in Number: (712) 775-7035
- Participant Code: 479307#
- International numbers: Conference Call Information
- Web Access: https://www.freeconferencecallhd.com/wp-content/themes/responsive/flashphone/flash-phone.php
Attendees
- Andrew Woods (DuraSpace)
- Bethany Seeger (Amherst College)
- Nick Ruest
- James R. Griffin III
- Andy Wagner
- Unknown User (acoburn)
Agenda
- We have consensus on the test plans
- Next step: execute via JMeter (early Fedora example)
- ...
Minutes
Consensus on Test Plans
Achieved consensus, 5 people have liked the Page)
JMeter Test Implementation
- Other actions is to put together the quite simplest JMeter tests
- (Just creating default containers)
- (More about the mechanics about dealing with JMeter)
- Some attendees of the calls have expressed some interest in the past
- Thus far, haven't heard from anyone regarding implementation
Utility within the Ongoing Development of Fedora
Some changes going on with the underlying Fedora implementation stack
- Modeshape is undergoing a major release next year
- Discussion of an alternate implementation
- Tests would assist in proving out the API
Baseline of tests now would be beneficial
A service (Blazemeter) can be hired (by DuraSpace) for JMeter test plans (executed in scale using, Amazon Web Services)
Progressing with JMeter
Question: Given the test plans...what are the appropriate next steps?
Examples from JMeter tests a few years ago
- Is it a matter of updating these configuration files?
- XML or GUI for creating the test plans
Experience regarding JMeter
- A. Wagner: New to Fedora...have used it on numerous occasions
- A. Woods also bears experience
- N. Ruest is quickly working to gain experience
- J. Griffin is a complete novice
- Would it be helpful to contact parties responsible for the legacy Fedora 3.x JMeter tests?
- The outlook is grim (due to the age of the work)
- Looking through commits...Nigel Banks, Chris Beer and Adam Soroka appear as contributors
- Also, there is not much a learning curve with JMeter...more a matter of time
- The limiting factor is time (more than anything else)
Looking to implement test case #4, one should allocate 30 minutes to try to open up the JMeter UI, load JMX file, have enough skill at create the tests
- Essentially, create the containers, and gauging by the speed of the response of the server, pass or fail the tests
- Ultimately, get to having JMeters for all 6 cases within the plan
- Then, decide what to do with the information
- Esme Cowles has been undertaking some testing
- Baseline of information drives the components of Fedora which must be adjusted in order to improve performance
- Note that the JVM configuration is included within the system profile
First Test Implementation Team
- N. Ruest leads
- J. Griffin volunteers
- A. Wagner is unavailable until 01/04/16
- Can look to offer hardware for building different boxes
- N. Ruest can spend next 1-2 days addressing this test
- Various test cases with different clients can be planned
- Incremental effort, where the initial creation of containers...building on itself for the container creation, and then look at the other tests
- 72 hours of N. Ruest's dedicated time will be invested in these efforts
- Following the updating of this Confluence Page following the call, N. Ruest will comment and update on the status of his labors
- Indicate any obstacles by Wednesday afternoon
- Should avoid forking the existing GitHub repository
- Include the JMX file alone into a new repository
Long Term Goals
Ideally, make substantial progress before February 2016
- Get a sense of how much effort it shall consume, produce a better estimate
- No strong commitments this far...
- Nice on the next call to have a better sense of when this initial set of tests shall be scripted out
Scheduling the Next Meeting
- Propose a meeting be scheduled into early 2016
- A. Woods shall be unavailable throughout the early year (no access to WWW)
- Hopefully have that meeting before A. Woods returns
- N. Ruest proposes 01/11/15 11:00AM (EST)
- A. Wagner authors an agenda Page
- Agenda will be filled in with updates regarding the progress with the tests
Outreach
- Reaching out to other parties which have, in the past, expressed interest in performance should be explored
- U. Michigan
- Hydra-in-a-Box at Stanford U.
- It would be ideal to bring in other groups (beyond the core Fedora committers)
4 Comments
Nick Ruest
I think I got a really basic container setup working with this.
I spun up fcrepo-webapp-4.4.0-jetty-console.jar, and poked around in JMeter until I got something. I let the test run for a while and filled up a fcrepo instance with a whole bunch of containers, and JMeter graphed something.
If somebody has some time to review and provide feedback, that'd be great.
James R. Griffin III
Thank you very much for the update Nick Ruest, I shall work throughout tomorrow and what remains of the week in order to test this.
James R. Griffin III
Nick Ruest I've successfully reproduced these tests within the fcrepo4 Vagrant Box (https://github.com/fcrepo4-exts/fcrepo4-vagrant) by installing and configuring a lightweight VNC server:
I can offer to focus next upon exploring the other test cases.
Andrew Woods
Nick Ruest, I have created a PR that:
https://github.com/ruebot/fcrepo4-jmeter/pull/1
James R. Griffin III, after Nick merges the PR, please reference this "Performance Requests" test fragment in your subsequent tests.