Re: storeGet() -> storeGetPublic() ?

From: Robert Collins <robert.collins@dont-contact.us>
Date: Tue, 9 Jan 2001 09:33:51 +1100

On the flip side I am seriously considering making some time for a database base metadata index. Probably using the sleepycat engine
or something similar. I *am* interested in trading space for time. (Because O/S swapping is not terribly smart.) On the other hand I
may pick up Bugless's memory mapped index and bring it up to HEAD level.

Rob

----- Original Message -----
From: "Adrian Chadd" <adrian@creative.net.au>
To: "Duane Wessels" <wessels@squid-cache.org>
Cc: <squid-dev@squid-cache.org>
Sent: Tuesday, January 09, 2001 8:04 AM
Subject: Re: storeGet() -> storeGetPublic() ?

> On Mon, Jan 08, 2001, Duane Wessels wrote:
>
> > I'm worried that you're heading down a one-way path.
> >
> > It seems to me your goal is to reduce the memory footprint,
> > and in order to do that you want to change storeGet().
>
> Nope. I'm not aiming to reduce the footprint. I'm aiming
> to allow different types of filesystems to be implemented.
>
> > Such a change is going to have an effect on performance.
> > Not keeping some of the StoreEntry bits in memory means it
> > will be slower to access them. You're trading space for time.
>
> Only if your FS goes out to disk to get the object information.
> For example, there isn't anything stopping someone from having
> a "ufs" filesystem which keeps the index in memory, and just
> populates StoreEntry's when they're needed?
>
> The problem is that squid isn't flexible enough to support some
> of the "alternate" object storage mechanisms. Why not try
> them out?
>
> > Some current Squid users may not appreciate being forced
> > into that tradeoff. For them, its better to have
> > a Squid that can serve 10,000 ICP queries per second
> > (and pay the memory costs) than it is to be stingy with
> > memory.
>
> Yup, I totally agree. Once I've completed my coding we can see
> what the results are. I'm hoping to be able to sustain at least
> that 10,000 ICP requests per second. :)
>
> Remember: I'm not forcing anyone to *just* use a disk based index.
> I'm simply trying to make things more flexible.
>
>
>
>
> Adrian
>
> --
> Adrian Chadd "Sex Change: a simple job of inside
> <adrian@creative.net.au> to outside plumbing."
> - Some random movie
>
Received on Mon Jan 08 2001 - 15:22:42 MST

This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 16:13:13 MST