https://homeostasis.scs.carleton.ca/wiki/index.php?action=history&feed=atom&title=DistOS_2023W_2023-01-09 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_2023-01-09&diff=24259&oldid=prev 2023-01-10T21:58:50Z

<p>Created page with &quot;Video from today&#039;s lecture is available on Brightspace, in Zoom-&gt;Cloud Recordings. ==Notes== &lt;pre&gt; Lecture 1 --------- What is a distributed operating system? First, it is an operating system - the code that transforms the computer that you have into the one you want to program - abstraction, resource management in the service of applications A distributed OS is one that runs across multiple computers connected via a network (or networks) So, why do we need...&quot;</p> <p><b>New page</b></p><div>Video from today&#039;s lecture is available on Brightspace, in Zoom-&gt;Cloud Recordings.<br /> <br /> ==Notes==<br /> <br /> &lt;pre&gt;<br /> Lecture 1<br /> ---------<br /> <br /> What is a distributed operating system?<br /> <br /> First, it is an operating system<br /> - the code that transforms the computer that you have into the one you want<br /> to program<br /> - abstraction, resource management in the service of applications<br /> <br /> A distributed OS is one that runs across multiple computers connected via a network (or networks)<br /> <br /> So, why do we need a separate class on this? Why can&#039;t this be covered in a regular OS class?<br /> - what&#039;s so hard about making an OS distributed?<br /> <br /> Networking is hard!<br /> - but why?<br /> <br /> State management<br /> - yes, but why?<br /> <br /> In a single-system OS, if we have a significant hardware or software issue,<br /> the system crashes/fails and we&#039;re fine with it<br /> <br /> In a distributed OS, individual computers can fail, be disconnected, etc<br /> - but the system should go on!<br /> - we have to, because at scale errors are common, cannot be avoided<br /> <br /> Failures make state a much harder problem<br /> - because when we try to synchronize state, it can fail, and then<br /> what do we do?<br /> - and, state sync is inherently slow over a network<br /> <br /> <br /> The problem of this semester is that generally-used single-system operating systems (UNIX-like, Windows, etc) use abstractions that are fundamentally incompatible with large-scale distributed operation<br /> - we can&#039;t keep the abstractions and also distribute the computation<br /> - we CAN keep using them, but only on individual systems<br /> &lt;/pre&gt;</div>

Soma