Transcript for SAS Containers
Ken Matz: (00:00)
If you’ve been reading about SAS Viya and you’ve been wondering how it deploys and whether it can deploy to containers, we’re going to be talking about those types of things today.
We’re here with Michael Koob. Michael, tell us a little bit about yourself.
Michael Koob: (00:15)
Thanks, Ken. I’m Director of Emerging Technology here at Zencos and as part of that role, I spend a lot of my time researching the newest of SAS’ products and solutions.
That includes SAS Grid but also includes SAS Viya, and of course being involved in the deployment space and config space for most of my time. We’re obviously very interested in the opportunities that container technology presents for the deployment of the SAS Viya product.
Ken Matz: (00:44)
Thanks, Michael. Tell us a little bit about SAS for containers and what it is.
Michael Koob: (00:49)
SAS for containers is primarily pointed towards SAS Viya, although SAS 9.4 is still usable in a container, so either solution can be there and containerized. So that’s what SAS for containers is, is it’s basically either one of those two products deployed into a Docker container.
Ken Matz: (01:08)
What type of business cases or main use cases do you see when you’re working with the customers that make them need to consider it?
Michael Koob: (01:17)
SAS in a container really is going to look to the end-user like the SAS they already know. So from an end-user perspective, whether or not the product is in a container or not really shouldn’t be material when it’s well implemented.
The focus of containerization is really around, the use case for the container itself.
And that is a much faster release cycle. Easier upgrades, rollbacks, and finally the flexibility of plugging it into, enterprise container workload managers like Kubernetes or say a proprietary, you know, binary to that like Open Shift and so many organizations are using that toolset, to more accurately control or software environments to manage the release cycle more effectively, maintain a faster, tighter release cycle and to better manage the density of their compute resources through the use of something like Kubernetes, to manage a hardware infrastructure as a monolith and keep the utilization of those resources high to maximize their value.
Ken Matz: (02:30)
For the listeners who don’t really know what Kubernetes is, can you tell us a little bit about Kubernetes?
Michael Koob: (02:35)
So Kubernetes originated with Google.
Google is the one that basically wrote the software. It is written as an open-source package. And so when I talk about OpenShift. Openshift is IBM’s proprietary version of Kubernetes, which comes of course with support. Open source communities have one available for use, but you would have to support it. [Which adds to the complexity.]
So there are a number of other vendors who have come into that space with Openshift being the leaders.
Ken Matz: (03:05)
Great. Thanks for that clarification. Are there examples of industries that we’ve worked with that you can share some stories about, some examples about what happened in those situations and then how SAS containers were able to help?
Michael Koob: (03:20)
We’ve been primarily seeing containers in the financial services space.
I think part of that is driven by the fact that they have larger research and development budgets than some other verticals. But that’s primarily where we’re seeing it is, they’re coming along and they’re starting to develop and have in house their Kubernetes.
You know, their infrastructure set up and as they start to move other applications or large parts of their applications into containers, they’re analyzing all of their applications and of course, so SAS is going to be looked at from that perspective as they review their applications and say, what can I put into a container?
And then you also have feeding back to the release cycle. You have, you know, for example, a SAS grid might only get upgraded once every couple of years. And the reason being is because it can be a labor-intensive process.
Michael Koob: (04:17)
And so I think a lot of organizations are looking at containers as a way to accelerate that release cycle more safely and more quickly, but newer products out there that their end users can use and for a much lower cost.
And so I think, the workload management aspect is a big one as well in that, containers properly deployed into Kubernetes, almost functions similar to a larger workload manager like LSF where you can intelligently pick hosts, based on their current utilization. But the work on to that host and then, you know, and kind of manages the workload for it.
So I’ve already got that. And so now I’ve got a workload manager that I understand and I have good support of and now I just need to know SAS.
I think it really fits well, it’s a leading edge platform. We haven’t seen a whole lot, um, in production just yet, but I think we’re very, very close to where folks are comfortable and confident enough with it that we’re gonna start to see some serious adoption in that space.
Ken Matz: (05:25)
It sounds really compelling, but what about the listeners or the customers who like the idea but they’re not sure or not competent that they have a container strategy ready to go?
What do you, what do you do in that situation?
Michael Koob: (05:39)
Well, I think that’s position that a lot of customers are in. The cloud has opened up a lot of different options and it seems like everybody is on some hybrid scale. And so that’s what we’re seeing from our customers.
We’re seeing that, you know, this software is being used in our research and development environments or they’re being put in as a proof of concept.
[Many] large financials that already have Kubernetes infrastructure. They’re doing proof of concepts, right? They’re trying to get these out there into there, Kubernetes, test them out and get ready to scale it out.
Then we also see, organizations that we work with that don’t have any Kubernetes infrastructure yet and we’re deploying there on just simple Docker hosts to prove the integration out, to again, let them do some playing with it and understand the capabilities integrated with the important other pieces, their environment, like data warehouses, Hadoop, things like that.
Make sure that authentication and other things that can be complicated by container, that all that stuff is working.
Again, this is all laying the groundwork. I see this as being something that gets very big in the very near future based on what we’re seeing in the field.
Ken Matz: (06:55)
So if they don’t have a fully thought out container strategy, that shouldn’t be an impediment because it sounds like they can do it together with the implementation of SAS for containers to further solidify.
Michael Koob: (07:06)
Yeah. Certainly it helps if they’re more advanced on that, but I don’t think from what I’ve seen, I don’t see that as a, there’s a clear line, this is the strategy.
So I think for somebody who doesn’t have a strategy, it’s important to get some thinking in that area, but I don’t think that puts that organization in necessarily a bad position because there’s a lot of that going on right now.
Ken Matz: (07:31)
That’s a good clarification. What key points or pitfalls to watch out for do you recommend to people considering SAS for containers?
Michael Koob: (07:39)
So SAS for containers. I think, one of the biggest differences that I see between it and other applications is the same one that we’ve always seen with SAS. If you want to deploy SAS in a container, in a Kubernetes infrastructure, you have to have hosts somewhere in that infrastructure that fit the resource demands that SAS has.
You can’t throw it at a bunch of hosts that have low IO capacity, poor networking, and talk to data sources, things like that. It needs to land on the right kind of hosts. But there still can be multiuse hosts or it’s, it can be a dedicated set of hosts that are related to SAS, but you still have the workload management layer in there to help you in Kubernetes.
I think it’s still SAS fundamentals driving, you know, setting up an effective infrastructure. Otherwise it integrates very nicely with the container.
Ken Matz: (08:31)
That’s a really good point to make sure that you still have the right hardware underneath before laying down the container layer. Before we leave here today, if you’re a customer and you’re liking the information that you’re hearing today, what do you suggest as the next steps to go towards thinking about and considering SAS for containers?
Michael Koob: (08:50)
So I think the obvious step is to find a partner that understands the technology and understands SAS in the long term, right?
It’s still a SAS in user that needs to work with the product. It still needs to deliver the same outputs, you know, if it’s not an end user-facing, this is batch programming.
You know, believe it or not, it’s for containers. It’s fantastic and batch processing. But you still have to understand what the system needs to do, the requirements, what the users have to do with it.
A partner that has a long history with SAS is going to be very beneficial. And then, of course, they need to know the newest technology that’s coming out of SAS, which is SAS for containers.
They need to have some experience with that product and the integration points that tend to be, you know, the sticking point of the implementations that containers ship.
Michael Koob: (09:46)
They ship in a very basic form in order to do the things like we’re talking about, Hadoop integration or operating system authentication, all those things, those need to be layered on after the fact.
There’s a lot of ways to do that. You want a partner that’s had enough experience to know what are the good ways, you know, what’s the best practices in those areas.
Really a partner that has a lot of experience with SAS, or they can be conversant with your requirements to meet for your users. And then who understands the technology well enough to deliver those requirements.