On May 9, 2019, at 4:56 PM, James K. Lowden <jklow...@schemamania.org> wrote:
>
> file I/O operations act on the kernel’s filebuffer
It’s more widely called a buffer cache or page cache:
https://en.wikipedia.org/wiki/Page_cache
> If this sounds like an indictment of NFS, it's really not.
In large part, it’s a restatement of some fundamental results in computer
science, particularly distributed computing:
https://en.wikipedia.org/wiki/Byzantine_fault
https://en.wikipedia.org/wiki/CAP_theorem
Reference 2 from the first Wikipedia article is especially worth reading:
https://www.microsoft.com/en-us/research/publication/byzantine-generals-problem/
Never mind the current host of that article: that just happens to be where the
2013 Turing Award winner and primary coauthor on this paper, Leslie Lamport,
was working when the Web took off. It was originally published in an ACM
technical journal while he was doing research at SRI International. In 1982.
That’s how fundamental this problem is.
As CS research papers go, “The Byzantine Generals Problem” is uncommonly
clearly-written.
That paper will explain to you why you want at least 3 servers with Bedrock if
you’re going to distribute it across multiple machines. It’s fine to start
with just 1 server, as long as you’re fine with a single point of failure, but
if you’re going to distribute the DB across machines, 2 makes no sense.
Intrigued? Read the paper.
Bedrock is based in part on the Paxos algorithm, another major advance in
distributed computing, and another of Leslie Lamport’s brainchildren:
https://en.wikipedia.org/wiki/Paxos_(computer_science)
_______________________________________________
sqlite-users mailing list
sqlite-users@mailinglists.sqlite.org
http://mailinglists.sqlite.org/cgi-bin/mailman/listinfo/sqlite-users
