On Fri, 12 Jul 2002, Kimbro Staken wrote: > >On Friday, July 12, 2002, at 12:53 PM, Heinrich Götzger wrote: > >> Think about the performance. XML-RPC brings a tenth of CORBA in >> throughput. >> >Fortunately, I don't think this is true. Last night I finally converted >Xindice over to using XML-RPC for all communications and running the >XML:DB API conformance tests usually falls between 18 and 22 seconds with >XML-RPC and somewhere between 16 and 50 seconds with the CORBA impl. I'm >not sure why CORBA varies so much, but I think it is related to the same >reason some people see really slow startups with CORBA at times. I was >seeing considerably slower times for CORBA when I ran the tests during a >high load time on my cable modem (i.e. 6-9PM), I suspect it has something >to do with nameservice resolution. The client and server were on the same >machine so the network shouldn't have been a factor otherwise. Debugging >this isn't something I really have any interest in doing since CORBA is >EOL. > >Anyway, these tests are not completely conclusive as I do need to run it >over the network, but initial results are promising. There's also still a >ton of room to optimize the XML-RPC impl because it is so much simpler to >know exactly what it is doing. In particular I need to finish implementing >compressed document support so that you don't have to parse the document >and query results again when they get to the client. We'll see how it goes. > Once I'm sure the API is more stable, I have another set of tests that >more realistically models a real application that I'll run to get a better >idea of the real performance. > >The code is in CVS if anyone wants to try it out, just be aware it >probably isn't very stable and I haven't evaluated resource consumption >yet.
This sounds promising. I'll wait untill a stable snapshot can be downloaded from CVS or per release to see how performance changes in my testbed. regards Heinrich -- http://www.xmlBlaster.org
