I always look at the programs for USENIX conferences. I’ve done this for so long that I can often pick out candidates for Best Paper awards just from the paper names and the list of authors. When I examined the LISA21 program, I was really excited, and found myself thinking about all of the useful tech that appears in this program.
I decided I’d ask the program chairs some questions about how they decided on this particular lineup. Unlike other USENIX conferences, with the exception of SREcons, there are no paper submissions to choose from. Instead, the chairs and program committee members have to come up with a program based on a list or proposals that they believe best fit the needs of their audience.
Rik Farrow: What were your goals for LISA21?
Carolyn Rowland: I'd like to see LISA continue to provide a place where IT operations folk can go to fill in gaps in their knowledge, learn what others are doing to solve challenging problems, and make connections with peers. I am hoping that the content brings people to virtual LISA because we can't draw them in with the good ole hallway track. I'm optimistic that the virtual platform will allow people to mingle in a new way as we navigate this strange socially distant world we find ourselves in currently. For those who miss the networking aspect of the conference, LISA is still here virtually until we can meet in person again.
Avleen Vig: Great answer! That covers most of my thoughts too. I would add that I've been passionate about the "Core Principles" track we launched at SREcon a few years ago, and bringing it to LISA has been a goal of mine. There's a rich opportunity here to provide education on core skills alongside the regular conference content.
RF: When I began learning about UNIX system administration in the mid-80s, most people were managing a single, mini-computer-sized system with lots of attached terminals. Managing disk space was a huge priority, as disks were small and users many. Today, I guess it is more likely that sysadmins today are managing instances in the cloud than local systems. Have things changed so much that it's more common to be administering to non-local systems?
AV: It's far more common, and it's not just cloud systems. Even if you have your own bare metal systems they usually live in some distant data centre you might visit once every few months. And then there's management of large numbers of remote devices, everything from set-top boxes to cars. We still have the same basic problems: how to handle resource constraints and how to manage resources. Now we have to do it across many systems, in many places, often with different sets of requirements. The more things change, the more they stay the same! I've been really happy to see the development of cgroups v2 into a mature system in the Linux kernel over the last few years that can really help us with some of these problems.
CR: I live in a completely different world of research devices. We have stuff in the cloud and bare metal servers but we also have robots, hand-crafted testbeds, a fire lab, HPC, etc. In one building I might be dealing with local HPC clusters, in another I have a large testbed built with the equivalent of 50 microwave ovens attached to a Windows PC running LabVIEW, in another building I have a variety of robots, while in yet another building I have a large fully instrumented burn lab trying to stream near-real time video across campus . Then over in a corner there's a special purpose machine tool controlled by a Windows 7 machine that can't be patched or upgraded. It's managing the chaos and making sure that research can happen despite the attempts at resource management and compliance. Every day is a new challenge. It keeps things interesting. So we have some modern systems and some "vintage" systems. We've got it all.
RF: I think things have really changed over time. In the 'old' days, core competencies were scripting, editing and communications skills — a distant third. Today, it seems to me that sysadmins and SREs will be using tools that support automation that make managing large numbers of systems even possible. cfengine didn't even exist until the early nineties, and that was just the start, and for revision control, we had SCCS, source code control system. How important are modern tools to current sysadmins?
AV: I think the toolkit that modern engineers have is invaluable. I can't imagine managing a fleet without it. Like a lot of tech-minded people I have a little server I use for a variety of things and even that is fully controlled by configuration management with every application running in a container. And all of those containers are built from data in source control. Everything can be rebuilt from scratch and backups restored in a few commands, and that can scale up from one host to hundreds of thousands of hosts without much effort. The scale and complexity of the systems and services we need to manage has increased exponentially, and without the tools to enable us to do this we would need armies of people to get the same work done.
RF: I see there are a lot of talks on SRE-related topics. Is this the direction LISA is heading, more toward what we would experience at an SREcon?
AV: This is a question we ask ourselves regularly: Where do we draw the line between the two conferences? There are a set of topics that are clearly good for SREcon, and another set that are clearly good for LISA. Then there's the fuzzy batch in the middle that could go to either conference — or both! A couple of years ago we decided that we wanted to direct topics that impact SREs (or Production Engineers, or Operations Engineers, or whatever you prefer to call them) more towards SREcon — this includes a lot of cloud computing, monitoring and observability, SLI/SLO discussions, and so on. That doesn't mean those topics would exclusively be at SREcon, but they are well suited to that conference. For LISA we wanted to bring in more Computer Systems Engineering topics — talks on bare metal systems, firmware, operating systems, kernels, systems management, data centers, security, and the wide range of things necessary to manage large installations. But that blend between the two conferences will always exist. In many cases people will find value from topics at both conferences in their daily work.
