[sr-dev] git:master:063e6a02: core: increase SHM memory pool to 128 MB
Daniel-Constantin Mierla
miconda at gmail.com
Wed Nov 27 10:28:42 CET 2019
Hello,
On 26.11.19 21:28, Henning Westerholt wrote:
> Hi Daniel,
>
> I don't think that anybody might think that Kamailio is greedy with resources.
If you start an application and demands a lot of memory without doing
anything, then people associated it with a resource greedy application,
like they did with Java or Electron based apps.
> A Raspberry Zero for 5 Dollar has 512 MB RAM and was released in 2015.
There are a lot of deployments out there that still run on old but
reliable (and expensive) hardware like IBM/HPE/Dell blades. You
comparison is not applicable, you look at the cheap consumer harware,
not what is actually out there in production environments.
>
> Some years ago we also increased PKG memory default for similar reasons,
At that moment there was a discussion about it debating the necessity:
commit 2376ef7c6f3317054c844ae9b99366d084065121
Author: Daniel-Constantin Mierla <miconda at gmail.com>
Date: Thu Oct 2 09:48:36 2014 +0200
core: default pkg memory size set to 8MB
- as per discussion on mailing lists
> so I think this is a reasonable change after all the years as well.
So it ends up being just your personal opinion, as I couldn't find any
evidence this change is actually needed. I do not find reasonable any
change to default allocated resources without proper grounds.
Not long time ago you published a blog post on high
capacity/performances of kamailio on raspberry pi, iirc, there was with
default config and no command line tunings, so I expect it was with the
defaults -- given that, 5000 registrations/second and 300 call setups
per second is more than what 90% of telefony systems deal with. This
change somehow contradicts it.
Cheers,
Daniel
--
Daniel-Constantin Mierla -- www.asipto.com
www.twitter.com/miconda -- www.linkedin.com/in/miconda
Kamailio World Conference - April 27-29, 2020, in Berlin -- www.kamailioworld.com
More information about the sr-dev
mailing list