Re: Metadata replication in tabled

[Jeff Garzik <jeff@garzik.org>]

On 06/28/2010 02:10 PM, Pete Zaitcev wrote:
> Well, yeah... I guess I was too lazy and reluctant to create yet
> another state machine for this. Maybe I should just bite the bullet
> and make tabled fully multi-threaded. It was likely to come next
> anyway since you complained about the abysmal performance (I do not
> know yet what the issues with performance are, but threads are
> likely to participate). But if so, a thread may just easily loop,
> as ncld API intends.


With my wikipedia corpus testing, part of the issue seems to be a lack 
of ready-and-waiting-to-write/read chunkd connections.  Ideally you want 
to have an already-open, initialized connection to a chunkd server ready.

Back in the days when I slung NNTP news across the galaxy, having 
multiple TCP connections to a single server was critical for 
performance, in particular mitigating the overhead of connection and 
authentication initialization.

	Jeff