Increasing the reliability of Horde Web-based email By Peter M. Abraham September 2003 Over the past month, our monitoring system as well as our customers have noticed our Web-based email (Horde) has been down on a frequent basis. We have been working diligently with our control panel vendor with the goal to increase reliability to 99.9% or higher up time. We've taken the following measures to increase reliability: We turned off error logging when error logging was not needed.  This reduced the processes running on the one server from over 400 to under 150.   We recompiled PHP using files instead of memory sessions; of note, we did notice the Horde email reliability problem start around the same time we switched to memory sessions from file sessions.   We are in the process of migrating approximately 60 Web sites to a brand new Parallels H-Sphere server from the existing Parallels H-Sphere Web server thereby lightening the load on the 1st Parallels H-Sphere Web server. For those of you who are studying RedHat Linux server administration, please note that the number of processes and server load are two separate indicators of the stress a server may be under at any one instance.  The server load on our initial Parallels H-Sphere web server has been typically under 0.75; it is now under 0.25 with the 1st two of the three moves. Should the above three steps fail to get us towards the up time percentage we desire, the next step will be to recompile the operating system kernel. Please note we at Dynamic Net take service, reliability, and performance seriously.  Thank you for your patience as we work on this very important feature offered by our company.