OpenLava

» see openlava.org as it lived, on the wayback machine

~ the goodbye ~

A cluster is a pile of machines and a queue of people who all want to run something on them, and a piece of software has to decide whose work goes where. In a great many research labs that software was LSF, a commercial scheduler Platform Computing grew out of the Utopia project at the University of Toronto in the early nineties. In 2007 Platform and Red Hat shipped a cut-down version of it under the GPL, called Platform Lava. Lava got one release and then nothing, and it was discontinued in 2011.

In June 2011 a Platform engineer named David Bigagli took the last revision of Lava and committed OpenLava 1.0 to GitHub. Compatibility was the whole proposition: the configuration syntax and the command line were left exactly where LSF had put them, so a site with years of accumulated submission scripts could switch and keep every one of them working. It was free and it behaved like the expensive thing, which is why places with more grant money than license budget picked it up.

In 2014 a group of former Platform Computing employees founded Teraproc and started selling support around it, with an Enterprise Edition carrying features the community build did not have. Platform itself was gone by then. IBM had bought the company in 2012, and the product OpenLava resembled was being sold as IBM Spectrum LSF. A free scheduler that behaved like the paid one, maintained by people who used to work on the paid one, was always going to attract a letter.

~ the chronicle ~

~ the letter ~

In October 2016 IBM sued Teraproc in the Southern District of New York for copyright infringement and misappropriation of trade secrets. The complaint said some of the company's founders had taken confidential and proprietary source code for Spectrum LSF with them when they left, and that this code had become the basis of OpenLava. On 17 October IBM's lawyers filed a DMCA takedown notice with GitHub. Bigagli posted that downloads had been disabled, and said he would pull the source for 3.0 and 4.0 and restore 2.2 in order to get the repository and the website back, while calling the claim fraudulent.

The case ran for nearly two years. On 18 September 2018 the court found for IBM and issued a permanent injunction against Teraproc and its agents. There was no shutdown announcement and no farewell post, only a docket entry.

~ what it says in the infobox ~

Wikipedia's software infobox for OpenLava has a field for the latest release version. The value sitting in it is "4.0 DMCA". Somebody updated the page after the takedown and wrote the reason straight into the version number, and it has stayed there ever since. The license field is equally blunt, noting that releases 3.0 through 4.0 contained confidential and proprietary source code from IBM Spectrum LSF. openlava.org itself returns a Cloudflare 523 with nothing behind it, and the redirect it offers over plain HTTP times out.

~ the part I can see from here ~

I work on edge AI and UAV computer vision in Ankara, which is the other end of this problem. The machine is small, it rides on the aircraft, and nothing queues. Batch scheduling belongs to the end where the compute sits in a room somewhere and your job waits its turn behind everybody else's. I never ran OpenLava. What is legible from this distance is the decision it was offering, which is the one every lab with a cluster has to make about whether to pay for the supported version.

~ what outlived it ~

Forks exist. The OpenLava 2.0 codebase was not named in IBM's complaint, so work derived from it survived the injunction: ByteDance maintains volclava, and a copy under one developer's GitHub account still carries the openlava name, though its last commit is from 2012. Neither is a continuation in any practical sense. What ended on 18 September 2018 was a scheduler with real users and real support contracts behind it.

@}-->--

              )
             (_)
            .-'-.
            |   |
            |   |
            |   |
            |   |
          __|   |__   .-.
       .-'  |   |  `-:   :
      :     `---'     :-'
       `-._       _.-'
           '""""""

« previous  ·  random  ·  next »