On 8/22/06, Sharon Yin <[EMAIL PROTECTED]> wrote:
The "serialize average time" I mentioned is the round trip time of one complete operation in our scenario, just like send a request from a client, handling inside servicemix, get the response and show in the client. In such round trip, the message has been through 7 BCs and 5 SEs, as well as 5 topic queue.
If your message weight 1 Mb and is serialized / deserialized 5 times, the servicemix performance won' t do much in this case. This is a pure jms problem.
Do you think it'll be more better for changint the flowName=jms to reduce the transformation time between Normalized message and jms message when going through the JmsSenderComponent and JmsInUsingJCABinding which publish and pull messages with JCA container? Does it matter?
Do you currently use the jms flow ? or are you asking if you should use it ? If you already use the jms flow, try configuring flowNames="seda,jms" The serialization cost may be prohibitive if you use the jms flow.
thanks, Sharon. ----- Original Message ----- From: "Guillaume Nodet" <[EMAIL PROTECTED]> To: <[email protected]> Sent: Tuesday, August 22, 2006 5:30 PM Subject: Re: Question about servicemix's run trip time > If you use 1 Mb size jms messages, you should try to minimize > unneeded jms roundtrips as much as possible. > > (Also, I hope you turned off debug logging, which would be killing > performances). > > I'm not sure how to consider the ratio of 1:3 you give. > It will mainly depend on what your components do. > The only thing you could really compare to, is when using jms > without ServiceMix. > > What do you mean by "serialize average time" ? > How is this parameter computed ? > > > > On 8/22/06, Sharon Yin <[EMAIL PROTECTED]> wrote: >> Hi,all, >> Does anybody have any idea about servicemix's run trip time? >> Recently, I've got some confusion about servicemix's performance. >> I implement some components as SE and deploy into servicemix, >> and configure some httpBC, JmsSenderBC, JmsInBinding BC, >> as well as using JCA container to fulfill our message driven architecture. >> we see that the result is not so good in our performance test, >> the run time proportion of our all components to servicemix's components is >> 1:3, >> which means that the servicemix's components run time takes 75% of total >> time. >> >> In addition, the max concurrent accessing number is 250 in our test >> environment, >> and serialized average time is 2.5s >> but when I use big message sized 1MB, the number reduce to 10, and >> serialized average time is 7s. >> >> I'm not quite sure if this is our components' coding low down the >> performance >> or it's because the frequent transformation between Normalized message and >> jms message takes too much time. >> >> hi, Guillaume, any comment about this? >> >> Thanks, >> Sharon. >> >> > > > -- > Cheers, > Guillaume Nodet > _________________________________________________________________ 免费下载 MSN Explorer: http://explorer.msn.com/lccn/
-- Cheers, Guillaume Nodet
