When we initially planned to migrate to LiteSpeed we thought this should take about a month. After we discussed how it should be implemented, LiteSpeed engineers developed a working version with dynamic vhost management in about a week which is really fast for such software. We installed it in our development environment and prepared migration tools to replace Apache.
Redis support was added to LSWS 5.4RC3. To our misfortune, we didn’t realize how many changes were planned and already done in LSWS 5.4RC3. The stable version release date was still unknown to us. RC4 was on the way too with additional changes that had to be tested. Here at Hostinger we often experience bad traffic coming to our clients` websites so it was a good environment to catch all unexpected errors. After the first deployment on our production server, we started receiving crash reports. At this point, bug hunting season started. For almost two months since the first deployment, we started reporting bugs to LiteSpeed engineers daily. Good thing is that when LSWS crashes for any reason, only a visitor that hit that bug is affected as LSWS generates a core file and gracefully restarts in almost no time. LiteSpeed engineers were focused on fixing those bugs and we got a new release in just a few hours after reporting them every day. When we stopped receiving crash reports from the first server we increased the server pool running with LiteSpeed, then new bugs started to appear. After about three months of such testing, verifying, reporting and fixing bugs a week came without any crash reports and we could tell that the stable version finally came.
After we deployed LiteSpeed, we saw a noticeable performance improvement compared to our old setup. Most of our clients use WordPress as their Content Management System so we tested it with LiteSpeed on our servers. We saw a great improvement on Time To First Byte (TTFB) compared to Apache.