Monday, 18 October 2010

How do you solve a problem like...

One of the most common initial comments from early adopters of iFind Discover is that they didn't see the benefit of having results from other institutions because they "only cared about items in their own libraries". They also expressed concerns that searches would take longer and explained that it was more difficult to determine quickly if items were available in their own library.

The issue of performance is something that we are only now beginning to look into as it has been impossible to determine where potential bottlenecks and problems might be with a limited number of people using the service. Searching three remote catalogues is bound to effect search time so we are keen to do some benchmarking to compare out installation with similar size single LMS implementations.

We initially experienced very poor performance during peak times which saw both our CPU and Memory allocation quick consumed. Luckily, as we running VuFind on a virtual server we were quickly able to add an extra CPU and double the system memory and things have been a lot more stable since.

We are still hitting peaks of 90% CPU usage during training sessions so we will continue to monitor the situation very closely. Possible culprits for these spikes are a) The use of Jangle during login b) The login workflow in general c) Server / client settings.

We took the comments from users regarding increased difficultly in getting item information very seriously and have tried to indicate exactly which institution availability information is coming from. The planned introduction of a new tab interface for holdings which clearly differentiates between each institution is on hold due to an Internet Explorer 7 bug.

Whilst we can take care of the ease of access to relevant information, the actual benefits of a shared catalogue will only be apparent if institutions enthusiastically agree to exchange resources and promote the service.


No comments:

Post a Comment