httpd-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Estrade Matthieu <estrad...@ifrance.com>
Subject Re: mod_cache CacheMaxStreamingBuffer
Date Thu, 12 Dec 2002 16:54:06 GMT
what i remember is cache is able to cache unknown size document with 
MaxStreamingBuffer.
i will try to draw it:

first brigade, no length, if no EOS bucket, saving data in temp brigade
second brigade, if no EOS, concat data to temp_brigade
until finding an EOS bucket, or total length is more than 
MaxStreamingBuffer.

All of this is starting line 630 of mod_cache.c

then, write_headers, and write_body (with temp brigade if the document 
is handled with more than 1 brigade), which are pointing on disk or mem 
cache writing functions depending on setting.

how long do you think the cache will stay in experimental
I am using it a lot in my product which is used by many customers
I found any big problems
in worker, reverse_proxy with mem_cache, i have like 300 request on 
backend for 120 000 request on reverse proxy
It increase performance a lot

regards,

Matthieu



Bill Stoddard wrote:

>>I confirm the CacheMaxStreamingBuffer is a source of misconfiguration
>>Using the MCacheMaxObjectSize could me a good thing.
>>In case of CacheMaxStreamingBuffer used, i don't know how
>>MCacheMaxObjectSize is used for because if a document is bigger than
>>MaxStreamingBuffer, it will be never cached.
>>are you sure MaxStreamingBuffer is not used in disk_cache ?
>>this directive is to store all the brigade before saving the entire data
>>with the write_headers and write_body functions, pointing on disk or mem
>>writing functions.
>>I think it's the same for disk and mem cache.
>>    
>>
>
>I need to look at the old code, but if i recall correctly, mod_mem_cache would
>reject attempting to cache an object whose length was unknown (see the call to
>cache_create_entity() in mod_cache). mod_disk_cache otoh would happily attempt
>to cache objects whose size was not know at the start. Remember that
>mod_disk_cache is just writing cache objects to disk and does not need to
>allocate memory to hold/prefetch the object.
>
>  
>
>>In the same time, could you answer my old mail with the
>>CacheSlashEndingUrl patch i made is usefull or not ?
>>    
>>
>
>Yea, I was just looking at that. I think we can eliminate that check. I was
>originally afraid that this would muck up negotiation but the cache does not
>work with negotiated content now anyway.
>
>Bill
>
>_____________________________________________________________________
>GRAND JEU SMS : Pour gagner un NOKIA 7650, envoyez le mot IF au 61321
>(prix d'un SMS + 0.35 euro). Un SMS vous dira si vous avez gagné.
>Règlement : http://www.ifrance.com/_reloc/sign.sms
>
>  
>


Mime
View raw message