Why Linux deployment could change everything
The release of Filemaker Server 19 with the new option of Linux deployment offers an alternative for business users which we have tested against the established market leader, Microsoft Windows Server 2016. But how good is it when compared with the Windows?
We have endeavoured to undertake a fair like-for-like comparison using the facilities provided by fmcloud.fm in France, Europe and on the eastern seaboard in the USA. We asked Claris Inc. whether we could also test their filemaker-cloud but that facility has not yet been made available, we hope to update this report with such a test in the future. We approached a couple of other hosting companies but did not receive any other offers of testing facilities.
Our objective was to establish what sort of viable alternative a Linux (CentOS 7.8) deployment would offer a business currently using a Windows Server 2016 service either on-premise or in-cloud? How would costs and performance compare?
How we tested performance
We tested using our open source dsBenchmark tool which was been widely used for load-testing FileMaker Server by many developers since 2015. It runs as a Filemaker App creating virtual user sessions using perform script on server on FileMaker Server. Claris engineers told us in 2016 that they load test in-house using a similar method. In load testing mode the number of virtual users is automatically increased until FileMaker Server becomes non-performant, thus establishing the deployment’s maximum potential user capacity which can then be compared with other deployments which have been tested in the same manner. Recently Wim Decort, who had assisted us in developing our research methods in 2015, released his own open source tool to do a similar job, Punisher. It will be interesting to compare our results.
A comparator not an absolute
This potential user capacity number is a comparator not an absolute because the reality of different business requirements, the efficiency or otherwise of their FileMaker systems, their hardware and their networks make direct comparisons difficult.