qpid-users mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "GS.Chandra N" <gs.chandra...@gmail.com>
Subject Re: SQL-92 subscription with Headers exchange on C++ broker
Date Wed, 18 Feb 2009 11:49:42 GMT
>On Tue, Feb 17, 2009 at 8:20 PM, Jonathan Robie <jonathan.robie@redhat.com>wrote:
>I honestly do not know which approach would perform better - we'll see! How
many conditions are you likely to be testing in a single query?

upto 100 conditions in 30% of the cases, upto 500 in 50% of the cases and
upto 1200 in the remaining cases.

The messages will be typically of 500 byte sizes with 5 headers on average.
The headers and values can be small, 4-15 bytes in size. The conditions will
involve headers and values of the same sizes. 99% of the conditions will
involve the same header in the format of the condition i have shown below.

A *minimum* of 500 clients is what will be required to be connected to a
single server.

Let me know if you need any other information. I will run some tests too and
post the results with many small subscriptions vs a huge few case scenario
in a few days time.

Thanks
gs

GS.Chandra N wrote:
>
>> That would be really cool Jonathan.
>>
>> The major reason i fear to take the multiple subscriptions route is due to
>> fears of performance degradation. If i can make do with a single
>> subscription, which performs better than equivalent multiple
>> subscriptions,
>> i could connect more clients to one server  (plausibly - still need to do
>> tests to figure out how many subscriptions can be supported by cpp broker)
>>
>>
>
> I honestly do not know which approach would perform better - we'll see! How
> many conditions are you likely to be testing in a single query?
>
>
> Jonathan
>
>
> ---------------------------------------------------------------------
> Apache Qpid - AMQP Messaging Implementation
> Project:      http://qpid.apache.org
> Use/Interact: mailto:users-subscribe@qpid.apache.org
>
>

Mime
  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message