[Eastern Time]
Additional Topics and Discussion
[Enrico Fermi Institute] 15:02:58
Yeah, So, have a few more things left in the slide that in the agenda let me share the screen again.
[Enrico Fermi Institute] 15:03:13
This was the the charge question that was posted before we had Katie's presentation.
[Enrico Fermi Institute] 15:03:20
The next thing was, you know, just kind of as we As we wrap up the the workshop, you know. Were there?
[Enrico Fermi Institute] 15:03:28
Were there, any, You know Other other topics, any other observations that folks wanted to bring up here?
[Enrico Fermi Institute] 15:03:35
I think it would be a great time to, you know, if there's anything that we haven't covered, that we should have covered. You know folks want to bring that up now. It's a great time
[Paolo Calafiura (he)] 15:03:47
maybe I missed it, and we discussed the about the next steps
[Enrico Fermi Institute] 15:03:51
Yeah, I have a slide for that. Actually.
[Paolo Calafiura (he)] 15:03:54
Good.
[Enrico Fermi Institute] 15:03:56
once you had something
[Lindsey Gray] 15:03:57
yeah, since Dale's still here, and I forget to ask it this morning.
[Lindsey Gray] 15:04:03
I wanted to understand. Going back to what you were talking about yesterday.
[Lindsey Gray] 15:04:08
I wanted to understand the process for setting up a level 2 or level 3 circuit, and if that just requires having you know, an appropriate certificate from an authority, and it can be, it can be largely automated or does someone actually have to physically appropriate things
[Lindsey Gray] 15:04:27
at some point in in whenever you're making every single circuit.
[Lindsey Gray] 15:04:32
How automated can that whole process be
[Dale Carder] 15:04:35
yeah, on the yes, that side for for layer 2. It can be fairly automated.
[Enrico Fermi Institute] 15:04:40
Okay.
[Dale Carder] 15:04:41
I forget off the end how it works, but that that's sort of tried true for layer through.
[Dale Carder] 15:04:46
That's to be defined. Typically, there's still other constraints.
[Dale Carder] 15:04:53
People might wanna consider such as is there adequate capacity on the physical infrastructure in which these services are virtualized.
[Dale Carder] 15:05:00
So there there still ends up being in like, you know, a a period of consultation before it gets the point of making dynamic.
[Dale Carder] 15:05:08
Api calls
[Lindsey Gray] 15:05:10
okay, yeah, as this will be a bit of a longer thought process.
[Lindsey Gray] 15:05:15
But I'm just wondering yeah, exactly how much of that can Billy really be automated?
[Lindsey Gray] 15:05:20
Because even in terms of like trying to understand what capacity you have at this, at the site or between 2 sites, that that's something that you could also figure out how to negotiate mostly automatically if you had some orchestration service dealing or dealing with both sites
[Lindsey Gray] 15:05:37
So huh? Alright, I'm gonna I'll chew on that some more. In probably send you an email.
[Dale Carder] 15:05:43
Yeah, also, and I think it's in the snow mass report on in the network section.
[Dale Carder] 15:05:47
I forget that was like Area 4. There's some some more stuff around that
[Lindsey Gray] 15:05:48
Okay.
[Lindsey Gray] 15:05:54
Alright great thanks.
[Enrico Fermi Institute] 15:05:57
Andy, you're in a comment
[abh] 15:06:00
yeah, there was a slide way back, I mean way back for today about what other services are needed to make cloud access.
[abh] 15:06:09
more efficient, better whatever, and I wasn't able to attend that particular section because I had a conflicting meeting.
[abh] 15:06:18
So I didn't want to mention a project of we're working on that that is called S.
[abh] 15:06:23
3 gateway. It was Big Thing is to be able to transition to the cloud easily by essentially allowing the current Hep security infrastructure, whether it be or tokens should be used with any cloud provider because with the S 3 gateway basically translates, all of
[abh] 15:06:46
those credentials applies the security that's necessary, and then does whatever you need to do in terms of uploading download to the cloud storage.
[abh] 15:06:54
Yeah. So that is in my mind, a a great way to branch the gap before we actually have to deal with the actual cloud provider security.
[abh] 15:07:07
so that people can start using this stuff relatively easily with their existing credentials.
[Enrico Fermi Institute] 15:07:16
Interesting. So I had a quick question about that. Andy does this only work in front of, you know a cloud provider?
[Enrico Fermi Institute] 15:07:24
Or would that be something that would work in front of Minio, for instance, that just provide Okay, awesome
[abh] 15:07:28
It definitely works in front of Menio
[Enrico Fermi Institute] 15:07:34
So does this help at all with the cloud certificate, authority type of problems?
[Enrico Fermi Institute] 15:07:38
Or is this
[abh] 15:07:40
not exactly. I mean, the problem is, it solves the problem perfectly for storage.
[abh] 15:07:46
Right doesn't quite solve the problem for compute access that you need something else for
[abh] 15:07:54
And I would assume Panda could handle that. But then, the panda theme would have a team would have to say something on that
[Enrico Fermi Institute] 15:07:54
Yeah.
[Enrico Fermi Institute] 15:08:03
But the compute is a bit easier, though right, because you know presumably the the experience, control, the environment that they run in so they can add whatever the ca authorities they want.
[Enrico Fermi Institute] 15:08:14
You know For instance, the if you, if you run the the Osg, you know regular grid.
[Enrico Fermi Institute] 15:08:17
C a cert bundle, or whatever that has. For instance, the in common.
[Enrico Fermi Institute] 15:08:23
I'm sorry, not in common, but the let's encrypt.
[Enrico Fermi Institute] 15:08:26
Ca: trusted right? So that that's a little bit easier.
[abh] 15:08:26
right, yeah, but I mean again it it It really depends on how that's integrated.
[Enrico Fermi Institute] 15:08:28
There.
[abh] 15:08:34
So I I don't know the specifics of on the compute side, and how that's done.
[abh] 15:08:38
So I can't really address that. But presumably, you know, like Panda, and equivalent systems, can provide the same kind of access
[Fernando Harald Barreiro Megino] 15:08:49
okay, for the integration with the cloud. What we do as we talk directly with the Google need to say Api: And But we have
[Fernando Harald Barreiro Megino] 15:09:02
Now. So this account Jason, into the harvesting machine, and and that's it.
[Fernando Harald Barreiro Megino] 15:09:10
Anything else.
[Enrico Fermi Institute] 15:09:16
Okay.
[Enrico Fermi Institute] 15:09:20
See, some hinders, Philip, did you guys have other comments or that's just cool.
[Fernando Harald Barreiro Megino] 15:09:25
no, I for
[Enrico Fermi Institute] 15:09:28
Okay? Yeah, done Okay?
[Douglas Benjamin] 15:09:31
Yeah, one thing I was wondering about is if
[Douglas Benjamin] 15:09:37
Right now at least for Atlas, the Hpcs and the Cloud Instances are what I'll call standalone in the sense of we are not flowing work.
[Douglas Benjamin] 15:09:54
We're not using either of which as an extension of our existing resources sort of a blending of on prem and off per in the same, let's say batch system for example, is that something we actually should be investigating for the future.
[Enrico Fermi Institute] 15:10:17
Are are you? Are you suggesting, like taking a tier?
[Enrico Fermi Institute] 15:10:19
2 and and adding cloud resources to it. And okay.
[Douglas Benjamin] 15:10:23
Something like that. I mean I know Dirk mentioned that they were sort of looking at that, and maybe it's not a tier.
[Douglas Benjamin] 15:10:29
2. Maybe it's it's in the analysis space.
[Douglas Benjamin] 15:10:33
Hello! What was shown with Reuben, though they're not doing on-prem off crime.
[Enrico Fermi Institute] 15:10:37
See.
[Douglas Benjamin] 15:10:42
In that they separated the science part to the cloud. Okay.
[Enrico Fermi Institute] 15:10:45
I mean. Personally, I think that in the analysis space cloud makes a lot of sense right?
[Enrico Fermi Institute] 15:10:50
Because you can get a lot of yeah, resources, any kind of resource you want.
[Enrico Fermi Institute] 15:10:54
You can get it very quickly if pay for it, obviously, but
[Enrico Fermi Institute] 15:11:05
I don't know if anybody else had other comments on that
[Enrico Fermi Institute] 15:11:13
Okay, other comments in general questions, other things, we miss
[Enrico Fermi Institute] 15:11:23
1 one. Meta question is, how well, did this format work for people
[Douglas Benjamin] 15:11:28
well, I have a question. There was a policy is that did that get Will that be talked about?
[Enrico Fermi Institute] 15:11:30
Yes.
[Douglas Benjamin] 15:11:36
Or is that you have a one line in the pilot on the old?
[Douglas Benjamin] 15:11:40
In the indigo that says policy, features.
[Enrico Fermi Institute] 15:11:43
yeah, I mean, I think in the, in the, in the code that was actually in the in the morning, right?
[Enrico Fermi Institute] 15:11:49
and that was just what what can we
[Enrico Fermi Institute] 15:11:53
You know one of those? What are the policies that we could ask for would make our lives easier
[Douglas Benjamin] 15:12:01
No cause. Some of the policies for the Hpcs can be rap can be very disruptive, and I know we didn't talk about it the Ls. The the incentives for the Lcs.
[Douglas Benjamin] 15:12:11
Right if we can work with doe and ultimately and that's for their biggest machines to really think.
[Douglas Benjamin] 15:12:19
Rethink the the policies that they, their incentives for what they, what there minders.
[Douglas Benjamin] 15:12:30
Hello! Them that might make might make these resources better for us
[Enrico Fermi Institute] 15:12:36
Yeah, yes, right?
[Douglas Benjamin] 15:12:37
Assuming we can run on them. Blah blah blah, So by that I mean as a very specific example, Argon Alcf schedules, the biggest jobs fastest from the smaller jobs, And yet, you know, we were able to atlas to you to run across the whole machine but we had to
[Enrico Fermi Institute] 15:12:59
Okay.
[Douglas Benjamin] 15:13:01
chop things up a little bit, which is a little different than you know the to different incentives.
[Douglas Benjamin] 15:13:10
And that's where we need help going up through Doe to talk 1 one office to the other.
[Enrico Fermi Institute] 15:13:17
Right.
[Enrico Fermi Institute] 15:13:21
Yeah, I mean, So how does that, you know? Get get translated into something that we We we put in the report right like we just say we you know we
[Douglas Benjamin] 15:13:30
So you do it as a finding, and then then maybe a recommendation.
[Douglas Benjamin] 15:13:35
If it's strong enough
[Enrico Fermi Institute] 15:13:36
Yeah.
[Douglas Benjamin] 15:13:38
That's how it gets translated into the report.
[Dale Carder] 15:13:38
ha! Having served on and say, ha! Having served on the review committees for, like some of the Lcs, it could be very interesting for you guys to read the annual reports.
[Dale Carder] 15:13:49
They write, for example, I'm sure Alcf has at least their 2021 report published, online and see the terminology they use, and how they describe.
[Dale Carder] 15:14:00
You know the essentially the the incentive structure they're given, and how they match jobs to the charge.
[Dale Carder] 15:14:06
They're given and phrasing things in those terms, I would certainly help
[Enrico Fermi Institute] 15:14:17
Yeah. This seems related to the comment that was made. I think, on Monday about.
[Enrico Fermi Institute] 15:14:22
You know we had to. We have to kind of sell it
[Enrico Fermi Institute] 15:14:27
In their terms
[Dale Carder] 15:14:28
yeah, it's so. And it's also there's you know, there's aside from just a difference between Htc.
[Dale Carder] 15:14:34
And Hpc. There's still sort of differences in language you know about how these workflows, you know, exist, and what resources they need, and why it should be a national imperative to make sure that.
[Dale Carder] 15:14:49
The resources are allocated for your Us. Case
[Enrico Fermi Institute] 15:15:02
other comments there
[Enrico Fermi Institute] 15:15:08
Other comments.
[Douglas Benjamin] 15:15:08
yeah, I think we phrase the language right and the reason, I say, that is because I had a conversation with the some activity I was doing Where That's where it came up.
[Enrico Fermi Institute] 15:15:20
Okay.
[Douglas Benjamin] 15:15:20
Good. If we can give the the Lc. Apps some coverage down, work with them to change the policy.
[Douglas Benjamin] 15:15:29
They really do want to have their machines be usable to a large cross-section of the science.
[Douglas Benjamin] 15:15:35
But they have as the way they're being judged against.
[Douglas Benjamin] 15:15:43
Until that changes they can't do things that make them much more usable for us.
[Douglas Benjamin] 15:15:48
So that's where we have to help them. Help us in.
[Enrico Fermi Institute] 15:15:50
Sure.
[Douglas Benjamin] 15:15:51
A report like this might be able to do that if it goes far, far enough up the food chain
[Enrico Fermi Institute] 15:16:03
Okay, Thanks. Time.
[Enrico Fermi Institute] 15:16:07
Al thoughts
[Enrico Fermi Institute] 15:16:17
Okay. Maybe I'll move on to the the last slide here.
[Enrico Fermi Institute] 15:16:20
So yeah, what happens next? I mean, first of all, you know, we we really appreciate everybody kind of slogging and out and doing this for for 3 days as a mix of mostly remote, and some some folks coming in person.
[Enrico Fermi Institute] 15:16:32
We really appreciate everybody's input, you know, I I think at this point, you know, our next steps would be to you know.
[Enrico Fermi Institute] 15:16:39
We'll we'll follow up with some folks after the workshop, I'm sure that will want to, pick your brains a bit.
[Enrico Fermi Institute] 15:16:48
More about specific areas of the report that we're writing.
[Enrico Fermi Institute] 15:16:52
yeah, we would really appreciate if you, you know, keep sending your feet back, keep sending us your thoughts on these topics, you know.
[Enrico Fermi Institute] 15:17:01
I I put some kind of tentative date here for a draft. You know.
[Enrico Fermi Institute] 15:17:06
We're going to plan to, you know, have a draft out by, you know, the first in November, and then our our report will be finished by December.
[Enrico Fermi Institute] 15:17:15
First. Yeah, Yeah, So thank you. You know any of the other folks had comments.
[Enrico Fermi Institute] 15:17:27
Stir, no no.
[Dirk Hufnagel] 15:17:33
No, I think I just wanna maybe add that, thanks to everyone that participated in the discussion, we are very appreciative of that was, we got some good discussion going now, we Just we have to go through the notes summarize get a report, maybe get more input If there's more input, follow up.
[Enrico Fermi Institute] 15:17:50
Yeah, yeah, the folks want to to help write pieces, you know, I think that's a that would be okay.
[Dirk Hufnagel] 15:17:51
With some people.
[Dirk Hufnagel] 15:17:59
There's still some hands up, so maybe there are some last minute comments
[Enrico Fermi Institute] 15:18:00
Yeah, indeed. Eric.
[Eric Lancon] 15:18:03
yes, I wanted to congratulate you, too, for organizing this workshop.
[Eric Lancon] 15:18:09
When I saw the Ad. I said I thought it would be mission impossible to, and bet you did it.
[Eric Lancon] 15:18:18
Oh, and it was very interesting, very fruitful discussions, Congratulations!
[Eric Lancon] 15:18:24
Thank you very much
[Enrico Fermi Institute] 15:18:25
Thank you. Tony.
[Tony Wong] 15:18:28
So you talk about keep sending us your feedback. Where should that feedback be?
[Tony Wong] 15:18:33
Just to send it to your own personal email is that how it works
[Enrico Fermi Institute] 15:18:36
Yeah, you know, I thought about that when I when I wrote that bullet point in the slide, I guess I should have put the put the email addresses on there, or put it's a mailing list or something Yeah, maybe Maybe we'll follow up.
[Enrico Fermi Institute] 15:18:48
With another mail, you know, with with how folks could could you keep contributing, and how we can keep the conversation going
[Enrico Fermi Institute] 15:19:01
Other comments.
[Paolo Calafiura (he)] 15:19:05
just want to echo what Eric said. Great job, guys
[Enrico Fermi Institute] 15:19:10
Hey? Thank you very much. Okay? Well, thanks. Everybody. Yeah, We'll follow up.
[Enrico Fermi Institute] 15:19:16
Appreciate your time.
[Lindsey Gray] 15:19:17
Thank you.
[Dirk Hufnagel] 15:19:19
hey? Thanks bye.
[Enrico Fermi Institute] 15:19:22
Hi! Everyone.