On Tue, Nov 20, 2012 at 4:32 PM, Zooko Wilcox-O'Hearn <[email protected]>wrote:

> If you're interested in distributed systems, please read and comment!
>

I'm interested in distributed systems! :D


> I posted an argument for why distributed, end-to-end, two-phase commit
> will probably work fine for LAFS's purposes even though it has gained
> a well-deserved reputation for "not scaling up to the Internet" in
> other contexts. (Hint: the answer is, of course, that we're demanding
> less of it than most systems do.)
>

I guess what I'm unclear on, both after reading this email and your notes
on #1755, is what algorithm you want to use for 2PC that you expect to
"scale up to the Internet"

I think a variant of Paxos is the go-to tool for this use case, and it is
most certainly known for scaling only to a small number of nodes.

Something I might propose for this use case: have you considered replacing
the introducer with something like Zookeeper (or at least supporting it as
an alternative), which can provide both a node registry and manage things
like distributed locking for you?

-- 
Tony Arcieri
_______________________________________________
tahoe-dev mailing list
[email protected]
https://tahoe-lafs.org/cgi-bin/mailman/listinfo/tahoe-dev

Reply via email to