W3C home > Mailing lists > Public > www-jigsaw@w3.org > May to June 1999

Re: cache policy

From: Yves Lafon <ylafon@w3.org>
Date: Tue, 22 Jun 1999 14:12:32 +0200 (MET DST)
To: Raymond Racine <ray@malema.com>
cc: "www-jigsaw@w3.org" <www-jigsaw@w3.org>
Message-ID: <Pine.GSO.4.10.9906221402530.13836-100000@tarantula.inria.fr>
On Mon, 21 Jun 1999, Raymond Racine wrote:

> I am inferring the you are referring to Jigsaw's LRU caching of resource
> meta information and not the resource itself.  This resource information is
> fundamentally some meta information that is part and parcel to the
> HTTP/1.0/1.1 protocols.
> 
> Does Jigsaw cache recently served HTML?  If I call the same HTML page thrice
> in succession does Jigsaw serve a cached (ignoring time out / expirations)
> HTML page each time or does it read in the HTML file via file I/O calls each
> time and therefore depends on the OS to buffer/cache disk I/Os.  I did not
> find this kind of caching of the resource itself (HTML file) inside Jigsaw,
> just the caching resource meta information (resource size, MIME type etc..)
> in the source code.

I was talking about the cache algorithm of... the cache (proxy cache).
For all the other caches present in the server, there is (roughly)
* a Client cache (the beast in charge of handling a request)
* a Thread cache
* meta-information cache (part of the persistent object, which stores all
  the meta information and the way to export views of the object, an HTTP
  request is a view of an object).
* a Client Connection cache (to handle persistent connections)
* a HttpConnection cache, for outgoing connection (proxy and such)

The client connection cache is the one with four parameters (see
properties->socket connection prop).

For the HTML pages, they are not cached, but the meta information is. I am
not sure that the time spent to handle a database of cached resources is
less than the time you save when you get the file from the disk, as on
modern OSes, the filesystem should do that and the mmap should get the
cached version (by the OS) directly, of course I suppose that a highly
demanded server runs on a good OS :)
Also, on SCSI hard disk, there is a memory cache, all this must be enough
to avoir more complexity ;)
Regards,

      /\          - Yves Lafon - World Wide Web Consortium - 
  /\ /  \        Architecture Domain - Jigsaw Activity Leader
 /  \    \/\    
/    \   /  \   http://www.w3.org/People/Lafon - ylafon@w3.org    
Received on Tuesday, 22 June 1999 08:12:49 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Monday, 9 April 2012 12:13:29 GMT