Random notes from a security-aware software engineer, open-source advocate and occasional lecturer.
When I was first trying out Nginx as a Maven repository proxy it already felt like Nginx was more performant and in any case less memory consuming than Nexus but with the systems running on completely different hardware, there was no way of getting accurate information about the speed difference..
That is why I set-up a test scenario with both systems running on the same infrastructure and doing several builds against the servers to get comparable.
For the test I used one of the Opencast Nexus servers and run tests against the set-up with the latest version of Nexus. Then I switched the configuration to Nginx with Nexus shut down and repeated the same test.
The machine I used for that was an AWS VM with 4GB of RAM and two cores. Nothing too powerful, but it should be enough for the test.
For testing, I then used the following script to repeatedly do a fresh build of Opencast with all local caches wiped clean:
for i in `seq 1 20`; do git clean -fdx rm -rf /home/lars/.m2/repository mvn -e clean install -Pdev | \ grep -A3 'BUILD SUCCESS' | \ grep 'Total time:' >> ~/buildtime.log done
The test was run at the University of Osnabrück on a fast desktop machine with a gigabit internet uplink over the Easter holidays, meaning that there was not much else going on on the campus network at the time. Optimal conditions to get data from the internet really fast.
For comparison, I also run the same test with everything cached locally and Maven in offline mode using:
for i in `seq 1 3`; do mvn -e --offline clean install -Pdev | \ grep -A3 'BUILD SUCCESS' | \ grep 'Total time:' >> ~/buildtime.offline.log done
Note that I did run three initial builds on Nginx to fill up its cache. The builds worked perfectly fine but were about 14min slower since all artifacts were requested from the projects main Nexus server. But that would be something Nexus would need to do as well on its first start.
This plot shows the build times for 18 builds of Opencast each with Nginx, Nexus and in offline mode. For the complete list of results, have a look at the data file.
While the difference is not devastatingly large, is is clear to see that Nginx is constantly faster. With an average time of 16:46min for Nginx and 17:37min for Nexus, there is a noticeable 0:51min average difference.
In this test Nginx showed that it definitely performs better than Nexus. Given that Nginx also needs less resources I suspect that under high load the difference would be even greater.
With these performance advantages and the reduced maintenance necessary for a tool packaged in all common Linux distributions, for me, Nginx is the clear winner and I will probably replace all except one Nexus server, on which features like uploading artifacts, … are actually used.