You are viewing a plain text version of this content. The canonical link for it is here.
Posted to user@jmeter.apache.org by Philippe Mouawad <p....@ubik-ingenierie.com> on 2017/10/23 18:19:36 UTC
Re: SampleSender modes
Hi,
Any feedback on this ?
Regards
On Thu, Oct 12, 2017 at 4:35 PM, Philippe Mouawad <
p.mouawad@ubik-ingenierie.com> wrote:
> Hello,
> We have currently 10 modes and I believe at least 3 of them are useless:
>
> - DiskStore : Probably overwhelms disks / IO and the fact that results are
> sent at end will probably induce an important network traffic and time to
> get the results. Any issue in network or injector will loose results.
> - StrippedDiskStore : The fact that results are sent at end will probably
> induce an important network traffic and time to get the results. Any issue
> in network or injector will loose results.
> - Hold : holds results in memory which makes it useless IMO
>
> I propose to drop them.
> If anybody uses them, let him speak now or remain silent until the end of
> times :-)
>
>
> Regards
> Philippe
>
--
Cordialement.
Philippe Mouawad.
Ubik-Ingénierie
UBIK LOAD PACK Web Site <http://www.ubikloadpack.com/>
UBIK LOAD PACK on TWITTER <https://twitter.com/ubikloadpack>
Re: SampleSender modes
Posted by Philippe Mouawad <ph...@gmail.com>.
On Mon, Oct 23, 2017 at 9:37 PM, Emilian Bold <em...@gmail.com>
wrote:
> These "modes" are actually "services". When you configure the mode in
> some .properties file you basically pick which service gets used.
>
> We should not have a SampleSenderFactory that does an `if` on modes,
> we should have a more generic mechanism where any such sample sender
> implementation might exist.
>
Agree, feel free to provide a PR.
>
> When looking at things like this, I don't see how a mode existing or
> not bothers anybody. Why shouldn't I be able to use a RAMdisk and use
> the 'DiskStore' "mode"?
So this would justify keeping Disk modes although user wouldn't have any
feedback on test until finished...
> Did you know Google Cloud has instances with
> hundreds of GB of RAM? Rather cheap too if you only needs if a few
> hours!
>
If you're speaking about Hold mode, what about the required JVM Heap and GC
impact.
I am not convinced it should be kept and have dropped it meanwhile.
But if you think it should be kept , I am ready to reconsider.
Still, I think we have too many options here.
> > Any issue in network or injector will loose results.
>
> Any reason JMeter server has to be so brittle?
>
Lack of time and contributions, not my priority until now....
> Instead of dropping modes, I'd say we focus on doing a sample sender
> infrastructure where bi-directional connections aren't mandatory (drop
> Java RMI) and allow some network hiccups (have some buffered message
> queues or something).
>
Yes, it has been discussed in the past, but no progress made.
I'd be more than happy to consider a PR.
>
> My 2c.
>
Thanks
>
>
> --emi
>
>
> On Mon, Oct 23, 2017 at 9:19 PM, Philippe Mouawad
> <p....@ubik-ingenierie.com> wrote:
> > Hi,
> > Any feedback on this ?
> >
> > Regards
> >
> > On Thu, Oct 12, 2017 at 4:35 PM, Philippe Mouawad <
> > p.mouawad@ubik-ingenierie.com> wrote:
> >
> >> Hello,
> >> We have currently 10 modes and I believe at least 3 of them are useless:
> >>
> >> - DiskStore : Probably overwhelms disks / IO and the fact that results
> are
> >> sent at end will probably induce an important network traffic and time
> to
> >> get the results. Any issue in network or injector will loose results.
> >> - StrippedDiskStore : The fact that results are sent at end will
> probably
> >> induce an important network traffic and time to get the results. Any
> issue
> >> in network or injector will loose results.
> >> - Hold : holds results in memory which makes it useless IMO
> >>
> >> I propose to drop them.
> >> If anybody uses them, let him speak now or remain silent until the end
> of
> >> times :-)
> >>
> >>
> >> Regards
> >> Philippe
> >>
> >
> >
> >
> > --
> > Cordialement.
> > Philippe Mouawad.
> > Ubik-Ingénierie
> >
> > UBIK LOAD PACK Web Site <http://www.ubikloadpack.com/>
> >
> > UBIK LOAD PACK on TWITTER <https://twitter.com/ubikloadpack>
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: user-unsubscribe@jmeter.apache.org
> For additional commands, e-mail: user-help@jmeter.apache.org
>
>
--
Cordialement.
Philippe Mouawad.
Re: SampleSender modes
Posted by Emilian Bold <em...@gmail.com>.
These "modes" are actually "services". When you configure the mode in
some .properties file you basically pick which service gets used.
We should not have a SampleSenderFactory that does an `if` on modes,
we should have a more generic mechanism where any such sample sender
implementation might exist.
When looking at things like this, I don't see how a mode existing or
not bothers anybody. Why shouldn't I be able to use a RAMdisk and use
the 'DiskStore' "mode"? Did you know Google Cloud has instances with
hundreds of GB of RAM? Rather cheap too if you only needs if a few
hours!
> Any issue in network or injector will loose results.
Any reason JMeter server has to be so brittle?
Instead of dropping modes, I'd say we focus on doing a sample sender
infrastructure where bi-directional connections aren't mandatory (drop
Java RMI) and allow some network hiccups (have some buffered message
queues or something).
My 2c.
--emi
On Mon, Oct 23, 2017 at 9:19 PM, Philippe Mouawad
<p....@ubik-ingenierie.com> wrote:
> Hi,
> Any feedback on this ?
>
> Regards
>
> On Thu, Oct 12, 2017 at 4:35 PM, Philippe Mouawad <
> p.mouawad@ubik-ingenierie.com> wrote:
>
>> Hello,
>> We have currently 10 modes and I believe at least 3 of them are useless:
>>
>> - DiskStore : Probably overwhelms disks / IO and the fact that results are
>> sent at end will probably induce an important network traffic and time to
>> get the results. Any issue in network or injector will loose results.
>> - StrippedDiskStore : The fact that results are sent at end will probably
>> induce an important network traffic and time to get the results. Any issue
>> in network or injector will loose results.
>> - Hold : holds results in memory which makes it useless IMO
>>
>> I propose to drop them.
>> If anybody uses them, let him speak now or remain silent until the end of
>> times :-)
>>
>>
>> Regards
>> Philippe
>>
>
>
>
> --
> Cordialement.
> Philippe Mouawad.
> Ubik-Ingénierie
>
> UBIK LOAD PACK Web Site <http://www.ubikloadpack.com/>
>
> UBIK LOAD PACK on TWITTER <https://twitter.com/ubikloadpack>
---------------------------------------------------------------------
To unsubscribe, e-mail: user-unsubscribe@jmeter.apache.org
For additional commands, e-mail: user-help@jmeter.apache.org
Re: SampleSender modes
Posted by Emilian Bold <em...@gmail.com>.
These "modes" are actually "services". When you configure the mode in
some .properties file you basically pick which service gets used.
We should not have a SampleSenderFactory that does an `if` on modes,
we should have a more generic mechanism where any such sample sender
implementation might exist.
When looking at things like this, I don't see how a mode existing or
not bothers anybody. Why shouldn't I be able to use a RAMdisk and use
the 'DiskStore' "mode"? Did you know Google Cloud has instances with
hundreds of GB of RAM? Rather cheap too if you only needs if a few
hours!
> Any issue in network or injector will loose results.
Any reason JMeter server has to be so brittle?
Instead of dropping modes, I'd say we focus on doing a sample sender
infrastructure where bi-directional connections aren't mandatory (drop
Java RMI) and allow some network hiccups (have some buffered message
queues or something).
My 2c.
--emi
On Mon, Oct 23, 2017 at 9:19 PM, Philippe Mouawad
<p....@ubik-ingenierie.com> wrote:
> Hi,
> Any feedback on this ?
>
> Regards
>
> On Thu, Oct 12, 2017 at 4:35 PM, Philippe Mouawad <
> p.mouawad@ubik-ingenierie.com> wrote:
>
>> Hello,
>> We have currently 10 modes and I believe at least 3 of them are useless:
>>
>> - DiskStore : Probably overwhelms disks / IO and the fact that results are
>> sent at end will probably induce an important network traffic and time to
>> get the results. Any issue in network or injector will loose results.
>> - StrippedDiskStore : The fact that results are sent at end will probably
>> induce an important network traffic and time to get the results. Any issue
>> in network or injector will loose results.
>> - Hold : holds results in memory which makes it useless IMO
>>
>> I propose to drop them.
>> If anybody uses them, let him speak now or remain silent until the end of
>> times :-)
>>
>>
>> Regards
>> Philippe
>>
>
>
>
> --
> Cordialement.
> Philippe Mouawad.
> Ubik-Ingénierie
>
> UBIK LOAD PACK Web Site <http://www.ubikloadpack.com/>
>
> UBIK LOAD PACK on TWITTER <https://twitter.com/ubikloadpack>