text stringlengths 0 1.15k |
|---|
And the other thing that we care about is exotic compilers. So we don't just build with GCC, we build with GCC, Clang, Intel compiler, Portland Group compiler, and then maybe the Cray compiler on the Cray machines... Lots of different ways to optimize your code. And there really wasn't a good package manager for experi... |
\[24:05\] Another problem with the software ecosystem here is that people will distribute their code, and maybe it relies on a very specific version of some research library, because maybe that research library is published like by a university, maybe they don't have the greatest release cycle set up, or maybe they don... |
Essentially, Spack is a tool to build software the way that I was building it, where I would have lots and lots of different versions of things, and essentially any dependency graph that you make is a new version, so we assign a hash to that. And that's a lot like -- if people are familiar with Nix or Guix or some of t... |
The idea is that you would be able to make a package that could build across all those different platforms and that people would be able to install it and say "spec install foo" and have foo, which is not generally the experience on supercomputers. |
**Nadia Eghbal:** Interesting just hearing some of the things that you have to be particularly concerned about, and it feels almost like you're doing open source but for a smaller but more engaged community of users, beyond your Lab. It sounds like a lot of it is for other labs, or that type of audience, more than a ge... |
**Todd Gamblin:** Yeah, I mean... I think for a lot of the simulation codes, the usage model is very different from what you see in web development. Depending on the team -- some of the teams are their own users, so it's a research project, they're developing this thing and they're testing out different versions of it ... |
Other teams do publish software that can be used by lots of people and they'll actually bother to package it themselves, but I think the build process on these machines is so complicated that your typical computational scientist doesn't wanna get into all those details; they wanna focus on the science. |
We actually talked to different teams in the community and try to figure out "What's your deploy process?" Some of them had good practices, and others said "Well, we have one user on every machine, because we build the code once, and there's this guy who runs it for everyone." Then everyone talks to him, they send him ... |
Now we're starting to see more sharing in the community, and also this sort of push for exascale computing has caused people to really think about how portable their code is... So things like this have started to matter to them. |
We have people who are really concerned with performance portability, which is another good reason to package your software and to use libraries that work across different architectures... For a long time - since like the '90s - if you could run on a Linux cluster you were probably okay and the processor architecture o... |
\[28:14\] A GPU is very different from a Xeon Phi in terms of how you would parallelize code for it, and it's very different from your multicore processor. So I think people have started to realize that they have to rely on libraries for that. They really wanna separate that concern out and give it to someone else who ... |
**Nadia Eghbal:** So how does that play into thinking about getting contributors for an open source project like Spack? Because I imagine -- I noticed that Spack has a lot of contributors, and there are only so many people that are using Spack in the first place, or your total audience is smaller... Do you feel like yo... |
**Todd Gamblin:** I think that one really intriguing place where we could get more contributors would be like industry... I think definitely we could grow the audience for it right now. I think there's a lot of people who still just build things by hand. |
I think when you look for contributors you only have to think about what's the structure of the community. For HPC actually I think there's maybe more roles than people are used to in the software community. I think you're used to thinking like developers and users, but actually we have -- so there's users, there's lik... |
Actually, one of the motivations for Spack was that that particular task was getting to be very overwhelming at Livermore computing, where I am. |
In my case - I was a researcher and I wanted to deploy things on the machines for students and post-docs who were working for me who don't necessarily know how to build all these things, but they want to run them. Then we also have a users support team that deals with the application teams, so they deploy software for ... |
The original contributors to Spack were people who were at these HPC centers who were just sick of building things by hand day in and day out, and wanted to deploy things on these machines. Our deployment model is a little different from what you might be used to from the cloud... Most of these machines have like a sha... |
So yeah, we targeted the HPC centers to try to get them to work together initially... But it turned out that we started getting more contributions from actual application developers and people kind of hacking on HPC code on their own, and I think that's actually served to really grow the community, but that wasn't some... |
**Mikeal Rogers:** The other HPC centers - are they as adept at open source as Lawrence Livermore is? Do they have similar policies around it, or are you a little bit ahead in that regard? |
**Todd Gamblin:** I don't know that we're necessarily better or worse than the other labs... I think there's strengths and weaknesses in terms of the process. There is no standard process for releasing open source software; it kind of varies from lab to lab. In terms of actually having popular research software, I thin... |
\[32:19\] I think in large part -- I mean, all the different laboratories have had some large open source project that they've put out there. I think DOE has had an open source software ecosystem for a while. I don't know that they've always had licenses or thought about the licensing aspect of these things, or thought... |
Accepting pull requests wasn't really easy before GitHub; even setting up infrastructure outside your laboratory could be difficult. A lot of Livermore teams have had trouble setting up all the hosting infrastructure that you need to host an actual project with collaboration tools, Subversion, things like that. So we'v... |
Al Livermore recently we've consolidated our GitHub presence, we've gotten more and more people to join the GitHub organization for LLNL; I think it's become easier, so more and more people are getting into that and really starting to think about how they put their software out there and how they do build communities a... |
**Nadia Eghbal:** I wonder how much - I know government is not a monolithic thing, so maybe just for you guys - of open source practices are being influenced by industry, versus they're doing it on their own... Do they care what companies are doing now with open source and looking at that and saying "Oh, we should do i... |
**Todd Gamblin:** I think different people are pushing that at different laboratories; it just really depends on the part of the laboratory. You could have two groups on the same hallway who feel very differently about this. We have some teams who are reticent to put their development version out there, they don't wann... |
Yeah, so we have people here who have really pushed to get -- Ian Lee, the guy who presented at GitHub Universe on Livermore's open source software, he really pushed to get people to consolidate there and to use open tools, to use sites like Readthedocs and stuff like that, and other labs have people like that as well. |
But you know, like I said, I think that labs have for a long time -- software projects, they don't think about building a community around it, and they don't think about it in the same way that the industry does now, in the way that I saw... Like, when I was at the Open Source Leadership Summit (the Linux Foundation ev... |
I wouldn't say though that historically the labs have been against that, I just don't know that they've thought about collaborating among the different labs as much. |
**Nadia Eghbal:** In terms of where you decide to start going with Spack, I noticed that Spack joined NumFOCUS as an affiliated project... For anyone who's listening, NumFOCUS is an umbrella organization for a lot of scientific and academic related open source projects, so I was wondering why did you decide to join Num... |
**Todd Gamblin:** \[36:25\] I don't know, it was kind kind of on a whim. I went to the NumFOCUS web page and I looked at their supported projects and their affiliated projects. I mean, I like all the stuff that NumFOCUS is doing; they're doing all kinds of awesome things, especially for the Python community, and R... |
I think that was right after Fernando Perez, the Jupiter guy (he's at Berkeley Lab); he had come out and given a talk here. So I don't know, I was inspired to go and do it. They had a list of requirements and they were like "You can be a NumFOCUS affiliated project if you this, this and this", so I wrote them an email ... |
And what does it mean to me? Well, I like what those guys are doing, and they said that if your project on the NumFOCUS web page, they can encourage people to contribute to your project. So I think for me it was that I'd like to be associated with this community so that potentially these other scientific developers cou... |
**Mikeal Rogers:** You've said something interesting earlier, which is that you've gotten more contributors from the user side of things than you have from the people maintaining the cluster side. Do you think that that's because you've democratized the whole role a little bit more, so you've made it easier and gotten ... |
**Todd Gamblin:** I think it might be a cultural difference. One thing I've found with the HPC centers is they don't adopt things easily; they have processes in place that they use to deploy software; that's been more of a socialization effort, like talking to them and saying "We're really behind this. This is a solid ... |
For example, one thing we did to get NERSC (which is Lawrence Berkeley Labs supercomputing center) on board is we worked with them to actually port Spack to work in the Cray environment. That was a fair amount of work, but they actually put in some developer effort and so did we, and we thought that was valuable becaus... |
I think one of the things that we did with Spack that was really helpful for getting more casual users to contribute was -- I mean, we looked at Homebrew and some of these other projects... I mean, Spack's package format is based on Homebrew, it's just Python and not Ruby. We looked at what we could do to make it reall... |
All you really had to do with Spack is clone it. It doesn't require you to be able to run Pip or some other Python package manager; you can just clone it and then you can run the Spack executable out of the directory there. |
I think that has helped get the regular hackers to start using it. That, and we specifically chose Python because Python is a popular language for scientific computing. I should say that Spack is not the first attempt to build an HPC package manager, there have been others. Oak Ridge had an internal package manager tha... |
\[40:07\] Actually, there's another popular package manager that's written in Python from HPC at the University of Ghent (that's in Belgium). It's called EasyBuild. They've done really good things for HPC packaging, but their tool is focused mostly on (I'd say) administrators of clusters, and it doesn't make it easy to... |
I was kind of modeling the Spack contribution model on Homebrew, because that seemed like a successful thing, and it's actually something that people in this community use on their Macs, and it seemed like they did a really good job of making it easy to contribute. So to a large extent, I think it was because we made i... |
**Mikeal Rogers:** Right, yeah. It sounds like the big differentiator for you has been focusing on the users rather than on the cluster maintainers. And, aligning with NumFOCUS is part of that, because most of what they work on is there; being in Python is definitely a part of that, too. |
**Todd Gamblin:** Yeah. |
**Mikeal Rogers:** Coming up, we get into funding for Spack and how Todd keeps this thing alive, what it's like working on a project from grant to grant versus ongoing programmatic support, and the challenges for open source in government, especially the Department of Energy. Stay tuned. |
**Break:** \[41:54\] |
**Mikeal Rogers:** So let's dig into it a little bit... Tell us a little bit about how you get funding for your particular projects. You mentioned a couple projects already that get you a fair amount of funding... How does the funding flow into your project specifically? |
**Todd Gamblin:** So for Spack, right now we are programmatically funded. What that means is we're funded by a program. I guess in DOE (and in government in general) a program is a giant source of funding that's been allocated potentially by Congress; that's kind of the best kind of funding, because it doesn't end, unl... |
There's lots of research funding in government and in DOE. We have an internal funding source, so for all the grants that we get at Livermore, we tax them and we have this thing called LDRD - I think a lot of the other labs do, too; it's called Lab-Directed Research and Development... So if you are doing research, you ... |
\[44:08\] There's competitive funding grants from the Office of Science, so that's another part of DOE; it's outside the NNSA, but we can apply for funding from there. In fact, a lot of our basic science research is funded by that. |
Then internally we have some funding pots that are discretionary... Basically, the management of the computational organization LLNL has discretionary money that you can go after. I think it's a more informal proposal process, but that's for things like hardening and maybe porting something to a new platform... Things ... |
Other things that people have been known to do - there's a whole SBIR program in the federal government... Are you familiar with that? It's Small Business Innovation Research. That is funding for small businesses to get started. A lot of things that started out as maybe programmatic projects and then sort of outgrew th... |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.