Re: Merge Window Closing Announcement for Core Update 127

[Michael Tremer <michael.tremer@ipfire.org>]

[-- Attachment #1: Type: text/plain, Size: 2548 bytes --]


> On 14 Jan 2019, at 13:32, Bernhard Bitsch <Bernhard.Bitsch(a)gmx.de> wrote:
> 
> Hi,
> 
>> Gesendet: Montag, 14. Januar 2019 um 11:43 Uhr
>> Von: "Michael Tremer" <michael.tremer(a)ipfire.org>
>> An: "Bernhard Bitsch" <Bernhard.Bitsch(a)gmx.de>
>> Cc: "IPFire: Development-List" <development(a)lists.ipfire.org>
>> Betreff: Re: Merge Window Closing Announcement for Core Update 127
>> 
>> Hello,
>> 
>>> On 13 Jan 2019, at 14:24, Bernhard Bitsch <Bernhard.Bitsch(a)gmx.de> wrote:
>>> 
>>> 
>>> 
>>>> Gesendet: Sonntag, 13. Januar 2019 um 13:43 Uhr
>>>> Von: "Matthias Fischer" <matthias.fischer(a)ipfire.org>
>>>> An: "Michael Tremer" <michael.tremer(a)ipfire.org>
>>>> Cc: "IPFire: Development-List" <development(a)lists.ipfire.org>
>>>> Betreff: Re: Merge Window Closing Announcement for Core Update 127
>>>> 
>>>> Hi,
>>>> 
>>>> On 13.01.2019 13:14, Michael Tremer wrote:
>>>>> I sent squid thousands of requests with ab (Apache Benchmark) and that worked fine with only one redirector.
>>>> 
>>>> Thanks for the feedback! Merged. ;-)
>>>> 
>>>> 
>>> 
>>> Nevertheless it's worth reasoning about multiprocessing ( #redirectors = #processors ) versus multitasking ( #redirectors = #task needed by the environment ).
>>> Even with only one processor it is possible to serve multiple requests. Using wait phases for other processes is a basic feature of multitasking.
>> 
>> This is not true for the redirectors though.
>> 
>> Our version of squidGuard takes only one request at a time. While processing that, it performs a database lookup (which is a btree) which is held in memory. So at no time it is reading anything from disk which would allow the kernel to interrupt the process and wait until the hard drive has responded.
>> 
>> Therefore it does not make any sense to run more than one process per CPU core.
>> 
> 
> But multitasking isn't accomplished by I/O-waits only. Task switches occur through time slicing and signals also.

Yes, but that is not a case squidguard is likely to run into.

Processing a request is fast and won’t take minutes or even seconds.

If the process is being paused and something else is running then there is no point in having multiple processes waiting. One process waiting is enough.

>> I guess update accelerator might suffer a little bit in the instance that it finds a match. Squidclamav is a different thing. It makes absolutely no sense to use that any more.
>> 
>> -Michael
>> 
>>> - Bernhard ( aka BeBiMa )
>> 
>> 
> 
> - Bernhard