https://homeostasis.scs.carleton.ca/wiki/index.php?action=history&feed=atom&title=DistOS_2023W_2023-03-20 2026-04-08T03:22:56Z Revision history for this page on the wiki MediaWiki 1.42.1 https://homeostasis.scs.carleton.ca/wiki/index.php?title=DistOS_2023W_2023-03-20&diff=24400&oldid=prev 2023-03-21T03:46:58Z

<p>Created page with &quot;==Notes== &lt;pre&gt; March 20 -------- Project Proposals - if you want to re-submit, you can, and I will replace the grade with your new one - Need to finish it by March 27th - some proposals were just way too broad - need to narrow it down - some were too focused on one paper - needs to cover multiple papers - some tried to argue for a technical point - you aren&#039;t doing original research - focus on what the papers say, have a thesis about patterns...&quot;</p> <p><b>New page</b></p><div>==Notes==<br /> <br /> &lt;pre&gt;<br /> March 20<br /> --------<br /> <br /> Project Proposals<br /> - if you want to re-submit, you can, and I will replace the grade with your new one<br /> - Need to finish it by March 27th<br /> <br /> - some proposals were just way too broad<br /> - need to narrow it down<br /> - some were too focused on one paper<br /> - needs to cover multiple papers<br /> - some tried to argue for a technical point<br /> - you aren&#039;t doing original research<br /> - focus on what the papers say, have a thesis about patterns<br /> in their arguments<br /> <br /> Remember the thesis comes after you&#039;ve read and understood the papers<br /> - represents some of the insight you got from the papers<br /> <br /> Borg &amp; Omega<br /> <br /> A container<br /> - packaged Unix-like userland<br /> - generally to be run on a Linux kernel<br /> - only dependency is on the Linux kernel, all<br /> userspace dependencies are included<br /> - popular in the cloud space because allows for composition of workloads on a host<br /> - can run without seeing each other with OS virtualization<br /> (private filesystem, user ID and process IDs, resource quotas, etc)<br /> - note that containers don&#039;t *really* contain from a security perspective<br /> - must use advanced OS security mechanisms (SELinux, system call filtering)<br /> to keep them separate, even then generally not enough<br /> - if running untrusted workloads, generally use hardware virtual machines to separate<br /> - area of my research how to do better! (look up bpfcontain, bpfbox)<br /> - remember, virtualization is not confinement!<br /> (virtualization is like putting on VR goggles, doesn&#039;t change how<br /> you can affect the world just what you see)<br /> <br /> <br /> Remember with Google (Borg, Omega), it is running containers that are trusted (i.e., they made them all), so they don&#039;t have to worry about containers going bad except if they are compromised, and they have other mechanisms to deal with that<br /> <br /> AFTER GROUP DISCUSSION<br /> <br /> What is Borg?<br /> - container orchestration<br /> - foundation for Google&#039;s distributed OS<br /> - unit of resource management is the container, not the process<br /> - notice the fate of UNIX, it isn&#039;t the operating system,<br /> it became the new process (rather than make a distributed UNIX)<br /> <br /> So they had Borg, why did they make Kubernetes?<br /> <br /> How did they make Borg?<br /> - Borg is old and dates from the early days of Google<br /> - didn&#039;t rely on hardware virtualization since their systems didn&#039;t support it!<br /> - so it is full of design choices that made sense at the time, but maybe don&#039;t now<br /> - for example, only one IP address per host<br /> - Borg was made specifically for Google, co-evolved with it<br /> - doesn&#039;t really make sense outside of their environment<br /> <br /> <br /> Kubernetes is the Google engineers getting a second chance to build Borg &quot;right&quot;<br /> - what they want to use<br /> - but really, it is a bit *too* general, can handle<br /> Google-scale workloads but very few need that much scalability<br /> and the cost is complexity<br /> <br /> What is Omega?<br /> - research project to do better scheduling than Borg<br /> - I&#039;m sure many of the ideas have been put into Borg by now<br /> <br /> Scheduling is a classic OS problem<br /> - lots of theoretical result<br /> - but in practice schedulers are quite messy<br /> - need domain-specific knowledge to decide what is important and<br /> to understand its needs<br /> - will tend to have heterogeneous workloads<br /> &lt;/pre&gt;</div>

Soma