I initially thought good option is to make use of the burden balancer to route all demands with ‘/rest’ towards the Relaxation API node. However, we found the Relaxation API seemed to be getting used throughout JIRA Software, such as the login page, so departing it to ‘/rest’ means we’d be mixing Relaxation API traffic with normal usage. We wanted the answer.
Dealing with a number of my co-workers, software development found we’re able to isolate Relaxation API demands by searching for ‘/rest’ in every request By searching at in which the request came from while using HTTP referer header. If an individual was attempting to login to JIRA Software or had been using JIRA Software, they’d get forwarded to or stick to an ordinary Usage node. Otherwise, when the person or bot was asking for the Relaxation API, they’d get forwarded to the Relaxation API node.
We suggested the reply to Cerner after a couple of models of testing, they went accept JIRA Software Data Center in October 2015.
The outcomes: performance at scale
Inside their first week of applying the suggested Data Center configuration, Cerner was seeing 4 occasions the quantity of traffic around the Relaxation API node as alternatively two nodes. Response occasions are faster, CPU utilization has decreased across their non-admin nodes in comparison one server instance, plus they haven’t seen just one unplanned outage in 2016, all while scaling JIRA Software to 1000’s of recent customers.
Cerner required to make certain that because software development ongoing to include customers that application reactions occasions maintained or enhanced. This re-architecting demonstrated that Cerner could reduce their response time by up to 50 %, from 150ms to 80ms. Even at peak traffic occasions – searching at page loads particularly – response occasions continued to be steady.
response some time and usage average response time pre and post using JIRA data center
To obtain the full story about how exactly Cerner scaly JIRA Software while enhancing response occasions and reducing CPU, listen to John and Mike themselves within our on-demand web seminar.