https://homeostasis.scs.carleton.ca/wiki/index.php?action=history&feed=atom&title=DistOS_2023W_2023-02-01 2026-04-08T03:25:31Z Revision history for this page on the wiki MediaWiki 1.42.1 https://homeostasis.scs.carleton.ca/wiki/index.php?title=DistOS_2023W_2023-02-01&diff=24331&oldid=prev 2023-02-01T19:11:04Z

<p>Created page with &quot;==Notes== &lt;pre&gt; LOCUS &amp; NFS notes ----------------- NFS: file servers &amp; file clients (few servers, many clients) - files live on the servers LOCUS: every computer has files, not all files are on all computers - so each computer could be a server or a client - or just have local access could have replicas but logically should act like one file - so replicas have to by synchronized - did centralized sync via a designated host - all updates would go there, and th...&quot;</p> <p><b>New page</b></p><div>==Notes==<br /> <br /> &lt;pre&gt;<br /> LOCUS &amp; NFS notes<br /> -----------------<br /> <br /> NFS: file servers &amp; file clients (few servers, many clients)<br /> - files live on the servers<br /> <br /> LOCUS: every computer has files, not all files are on all computers<br /> - so each computer could be a server or a client - or just have local access<br /> <br /> could have replicas but logically should act like one file<br /> - so replicas have to by synchronized<br /> - did centralized sync via a designated host - all updates would go<br /> there, and then it would distributed the updates to the replicas<br /> <br /> Is LOCUS very fault tolerant?<br /> <br /> files could only be as big as any disk<br /> - really, idea was small files, not large ones<br /> <br /> Partitioning was a concern, they have recovery methods, but really individual hosts going down would not lead to good performance<br /> - would try to recover, but fundamentally fragile<br /> <br /> LOCUS tried for transparency, but faults would break that model fast<br /> NFS did too, mostly worked, except with faults or if you unlinked open files<br /> &lt;/pre&gt;</div>

Soma