Capacity Planning
This article is relevant to clients running self-hosted instances of Elements. Symplectic manages capacity planning concerns relating to cloud-hosted instances of Elements.
In order to determine the specifications for your IT infrastructure that will facilitate the installation and implementation of Elements we have to consider the size of your institution and how your staff intend to use Elements.
Within this article we have collected some useful information and guidance - this can be used by staff members tasked with planning the capacity of the Elements SQL database. If your infrastructure environment is complex and highly customised you may need to seek additional support to determine the appropriate configuration. Please don't hesitate to contact the Implementation Manager at Symplectic or log a support call and we will involve our subject matter experts to assist you.
IOPS - Database Disk Performance Requirements
Peak IOPS depends on the follow factors:
The level of activity in Elements.
How you configure Elements. Many of the rates/schedules used by Elements are configurable.
How many concurrent users are using the system.
How many scheduled jobs are running.
How many systems you have connecting to the Elements API.
As a guide, during internal testing the heaviest component tends to be the reporting synchroniser where peak IO of around 100MB per second has been recorded.
Database Size And Growth
Listed below are some factors that will determine the initial size of your Elements application database.
Number of publications.
How many publications are matched to external data sources.
Number of researchers.
How many researchers use Elements.
Discipline coverage in Science, Technology, Engineering and Medicine (STEM) results in more matches to multiple external data sources.
Number of publications with extremely large co-author/collaborator relationships Example: High Energy Physics (HEP).
Number of subscriptions to external data sources at your institution.
The interplay of relationships that gets created when many of the above factors are present will also affect the growth rate of the database, along with the total number of research outputs produced by your research staff each year. To give you some data points:
An empty Elements database takes up around 61MB
An Elements database with around 50 active users, 15000 objects and 40379 records (average of 2.7 records per object) takes up around 1.5GB
An Elements database with around 6k users, 0.5M objects and 1M records takes up around 46GB
The large database example above (which is based on real data from one of our partner institutions) gives an average storage per object of 0.13MB per record. This figure is likely to be an overestimate as it is primarily based on publication data which tends to be both larger and richer than data for other categories.
We suggest that the above is taken as an indicative figure and multiplied by the number of publications (e.g. journal articles) currently associated with your institution for an estimate of initial size and by the average number of publications produced per year for an estimate of growth rate.
You should, however, factor in extra scale if you are planning to make use of other modules such as professional activities or grants.