Page 5 of 31
Colleague Self-Service Hardware and Software Configuration Guide
© 2016 Ellucian. All Rights Reserved - Confidential & Proprietary
Software capacities and limitations
Each component within the Colleague Self-Service stack has upper limits on the number of
concurrent users or concurrent request that can be handled before performance begins to
severely degrade. Many of these capacity limitations are based on how the software is tuned
but a few are based on hardware and architecture of each component. The numbers given in
this section are based on the tuning guidance within this document with the individual
capacities/limitations of each component taking into account the other supporting components in
the Colleague Self-Service stack.
Colleague Self-Service
A single instance of Colleague Self-Service can handle around 700 concurrent users/page
requests before the web application begins to degrade in performance. Short bursts can be
tolerated but sustained load above this number will result in noticeably slower response times.
The capacity is limited primarily by the number of concurrent connections allowed to the
Colleague Web API (Colleague Web API Connection Limit setting). Colleague Self-Service is
primarily an I/O-bound application that relies on the Colleague Web API for all data I/O. As
such, the number of concurrent API requests allowed has a large impact on how long it takes to
service a page request. To a lesser degree the capacity is also limited by the number of open
sockets on the host OS as browsers and web application like Colleague Self-Service open
multiple persistent connections to facilitate the application functionality. The large number
sockets opening and closing takes away CPU cycles for running applications.
Colleague Web API
A single instance of Colleague Web API can handle around 400 concurrent API data requests.
Short bursts can be tolerated but will result in increased response times. The Colleague Web
API is a CPU-bound application as the majority of the workload is reading/invoking transactions
in Colleague and transforming that data into JSON responses. Besides CPU, the capacity of the
API is greatly limited by the number of concurrent connections to the DMI application listener
(Connection Pool Size setting) that can be used to interact with Colleague and the amount of
time the slowest transactions take. In order to help limit the amount of I/O performed against
Colleague, the Colleague Web API caches source data when possible where the source data
does not change often or at all under most operational circumstances. Even with caching in-
place, the number of concurrent connections allowed to the DMI application listener and to a
greater point, the speed of the transactions has a large influence on the overall performance.
Take registration, for example, which is traditionally one of the slowest transactions. As more
registration transactions consume the allotted concurrent connections to the DMI application
listener, all transactions (fast or slow) have to wait for the resources to free which has a
snowball effect that causes all request to slow.