Freeradius with JRadius Performance issues.
Hi, I am evaluating freeradius with jradius. Currently i am performing performance testing for the Freeradius-Jradius combination using radperf tool. JRadius is simply authenticating the user from the jradius-config.xml file using one of the default handlers, so nothing fancy there. Here are a few numbers that i get. Standalone Freeradius 500 requests/sec gives me 461 1000 requests/sec gives me 858 Freeradius with JRadius ( Standard 8 connections in the configuration) 500 requests/sec gives me 60 1000 requests/sec gives me 152 So, as you can see, the performance numbers are way lower when JRadius is used. I tried increasing the connections and the threads in jradius but it did not help. I understand that since the freeradius to jradius connection pool uses tcp sockets, the performance will be lower than when freeradius is run by itself but this low is something I did not expect. Has anybody run into this issue? Is there a way around to getting the performance numbers higher? Thanks for your help! Hema.
Hemlata Shekatkar wrote:
I am evaluating freeradius with jradius. Currently i am performing performance testing for the Freeradius-Jradius combination using radperf tool. JRadius is simply authenticating the user from the jradius-config.xml file using one of the default handlers, so nothing fancy there. Here are a few numbers that i get.
Standalone Freeradius ... 500 requests/sec gives me 461
Hmm... that seems a little slow. What kind of tests are you running? Authentication? Accounting?
Freeradius with JRadius ( Standard 8 connections in the configuration)
500 requests/sec gives me 60
That's likely due to the large overhead of sending data back && forth to the JVM.
Has anybody run into this issue? Is there a way around to getting the performance numbers higher?
Try using radperf with "-p 500" rather than "-n 500". That will measure the maximum packet rate that the server can handle. Alan DeKok.
Thanks for the reply Alan. 1) I was running Authentication tests only with Radperf. 2) I ran the tests with -n 500 and the results were really awful. Only 15 for the 500 requests sent. I have run tests with both -n and -p options with various request numbers and the results with -n have been really bad consistently. I also increased the memory for jradius to xmx512m but still did not get any higher performance or load numbers. Any ideas? Thanks Hema.
Hemlata Shekatkar wrote:
2) I ran the tests with -n 500 and the results were really awful. Only 15 for the 500 requests sent.
That seems to indicate that the jradius portion has *extremely* long latencies.
I have run tests with both -n and -p options with various request numbers and the results with -n have been really bad consistently. I also increased the memory for jradius to xmx512m but still did not get any higher performance or load numbers.
Email me off-list, and we'll see if we can figure this out. Alan DeKok.
participants (2)
-
Alan DeKok -
Hemlata Shekatkar