Full Speed Ahead
Following the launch of KHIS 2.0, the biggest gripe reported by users was the slow system speed. Upon investigation, this was found to be due mainly to the large amount of data imported from KHIS 1.5.
During the testing phase only recent data (i.e. less than two years old) was imported from KHIS 1.5. This volume of data did not cause significant speed issues and was not reported as being problematic during the user testing phase. However, upon the launch of the live system in November last year, all KHIS 1.5 data was imported. It dated back as far as 1995 and comprised several hundred thousand records; only at this point did the speed issue become apparent.
To combat the problem, several actions were taken:
- Archive old data
I.e. projects and all related data (e.g. documents, emails, actions and financials).
Due to the volume of data, it was not possible to simply bulk archive. Instead, key users from each of our five member universities (Durham, Newcastle, Northumbria, Sunderland and Teesside) were asked to provide a set of “archive criteria” which could be applied to data owned by their institutes.
Typically, this was projects that had a status of “closed” or had been inactive for two years.
Once all old data had been archived, system performance notably improved. Functionality was then added to the system to allow archived data to be searched if required. By default, though, searches are set to only search non-archived data.
Subsequent feedback from users confirmed that system performance had improved. Involving the users in addressing the problem subsequently led to a greater buy-in in KHIS 2.0. By actively involving them, the users became more understanding as they could see problems were being proactively tackled.
- Swap server roles
The KHIS 2.0 database was hosted on a dedicated database server. Similarly, the web site was hosted on a dedicated web server. However, the database server was several years old, whilst the web server was brand new, and a higher spec.
As SQL Server is more resource intensive than ColdFusion (the language KHIS is written in), it was decided to reverse the roles of the servers. This way, greater processing power and more memory would be available to SQL Server, which would therefore improve performance.
The server roles were swapped in several stages and, once everything had been implemented, performance was found to have improved even further.
