Re: Much wanted feature. (fwd)

On Tue, 28 Oct 1997, Dancer wrote:
> I'd suspect you missed by an order of magnitude. I'd guess at 1000 to 2000GB,
> myself. This is (of course) a Wild-Assed-Guess(tm). YMMV.

You're probably right, a big-chunk cache box would probably need a
terabyte disk array. Hmmm. Digital makes one of those, maybe NLANR could
get one for a root cache or two... It would be neat to see the hit rate
attainable if disk space wasn't [much of] an issue.

> Content-type: verbiage/waffle ; boundary=end-of-waffle

Waffle, eh? Havn't heard this one before... Must be a new MIME type :-)

> Tried it via NFS over 10MB ethernet. Acceptable. No noticeable delays from the
> client end..I mean, heck...The client's probably connected at 28.8K if you're
> an ISP. That's about 3.6Kb/s (taking compression into account) at peak. You
> don't need a whole lot of responsiveness out of a network mounting to keep up
> with those. Even a lot of those.

I would suspect it would work better on a network where collisions wern't
a problem, such as token ring or FDDI. Maybe switched Ethernet.

> Personally, I feel that there must be a better way to do things. But then, I'd
> be keen on seeing a persistant TCP connections between
> siblings/parents/children down which ICP requests can be funneled. I _like_
> the whole ICP thing...I just hate seeing it fall apart just because things get
> busy and the UCP/ICP packets get lost.--end-of-waffle

I would like to see it to. I think that's one of the things that's coming
in the long-awaited Squid 1.2. I sure do hope it's one of the additions to
Squid anyway!

        -Bill
Received on Mon Oct 27 1997 - 16:37:32 MST