Re: icp bitrot?

From: Robert Collins <robertc@dont-contact.us>
Date: 09 Oct 2002 08:39:29 +1000

On Wed, 2002-10-09 at 01:57, Duane Wessels wrote:

>
> > Actually, I'm creating async storeGetPublic* functions. This requires
> > breaking every synchronous function that use storeGetPublic up into a
> > start and completion routine. ICP uses storeGetPublic...
>
> What will this do to ICP performance?

Nothing at the moment. If/when we move the store hash table out to each
cache dir , and thus possibly to disk, then icp latency will increase.

However, one thing I've been thinking about is allowing a 'memory only'
flag to the new getPublic* calls, to indicate we are in a latency
critical section, and only in-memory key tables should be queried. I'm
not convinced that we'll need that though.

Rob
 

Received on Tue Oct 08 2002 - 16:39:33 MDT

This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 16:16:54 MST