We recently experienced a degradation in peformance affecting GXP Manager. During this event users of GXP Manager may have experienced slow loading times when using the product.
GXP usage varies throughout an average day and sometimes load can increase or decrease in an unpredictable manner. We continuosly monitor the performance of the plaform, and in cases where our tools detect an increase in load, we automatically provision more resources on a component by component basis.
The thresholds configured to trigger these scale events were in this case too conservative, this in turn caused scaling to be triggered too late. This caused a gap of about 30 minutes where response times was no longer acceptable.
This is indicated by a 30 minute window where the APDEX of GXP Manager dropped from 0.98 (Satisfied) to 0.87 (Tolerated).
Customers could still use the product but some operations could take considerably longer than normal.
After investigating we found that in some scenarios the threshold used to trigger a scale event was too conservative. We have altered this value to ensure that scaling happens sooner.
Scaling in itself can cause a temporary performance degradation. This is due to requests being temporarily held back while the scaling takes place. It is therefore a balancing act to ensure that scaling is only performed when needed.
We believe that the new setting will reduce the risk of another event of this kind.
We apologise to any customers that were affected by this event.
Feel free to reach out to support@iris.net should you have any questions or concerns.
Kind regards,
iRiS Systems Engineering