Performance At Scale: Can JIRA handle 10 million issues?

Insight Main Image

The following article is a guest post to Zephyr from Stefanie Chernow, Content Manager at Valiantys. Valiantys is a partner of Zephyr and empowers teams and enterprises to transform the way they work with the most powerful Agile tools on the market – the Atlassian suite.

JIRA is becoming mission critical for many companies - if JIRA goes down, it takes everything and everyone with it! Yet the more your organisation scales, the more you rely on JIRA project management and ironically the more you need to worry whether JIRA has the capacity to handle your success. Many people are starting to wonder at what point JIRA will deteriorate.

More and more we find that agile teams are asking the same questions: When do I need to switch from Server to Data Center? What is the threshold? How should I anticipate this?

Does Data Center even have the capacity to manage the volume that I am forecasting?

The aim of our study "Performance at scale: Can JIRA handle 10 million issues?" was to start exploring JIRA's tipping point and to reassure our clients that JIRA Data Center could handle their scalability.


Our environment consisted of a four-node Data Center with a relatively basic number of configuration objects - then we loaded it with 10 million issues. Our configured was as such:








Custom fields













Apache JMeter was used to simulate the behaviour of 2,000 concurrent users with four, three, two and one nodes.

Of course, every study has its limitations; larger companies with more complex configurations and several add-ons will get different results than our conclusions. Additionally, we used Amazon Web Services for the hosting, however the results would be different if another option was used.

However, this model mirrored an actual customer environment, so we've found that our results are a valid starting point for knowing JIRA's tipping point. In upcoming studies, we hope to further address some of these limitations.

Results: Data Center performance in a nutshell

We measured the average time it took to perform common actions that are regularly executed by JIRA users. As you'll see, there was a significant performance improvement from moving from one to two nodes. At three nodes performance was optimised for this environment. Four nodes didn't really provide a significant improvement over three nodes, however it still does allow for lower error rates and better response times.


Average time (milliseconds)


Additionally, there were significant improvements to how the CPU was used - which is key to JIRA's responsiveness and stability - by using Data Center. During the test we saw that CPU capacity was one of the main performance bottlenecks: as we lowered the number of nodes from four to one, the application CPU usage exceeded 90% and JIRA became unresponsive to users. Note that during the single node tests, the CPU reached its capacity, causing a very high error rate (28%) with a large number of requests being dropped. In comparison, with four nodes, CPU usage never exceeded 50%, and the error rate was at 0.07%.

On the other hand, the memory usage was not affected much when we allocated 21GB for each node (utilisation never exceeded 13GB).


Overall, the study concludes that there is a significant boost in performance between a single node architecture on JIRA Server and JIRA Data Center's powerful multi-node cluster architecture - JIRA admins can be up to three or four times faster!

However, again every environment will be different and this study is just the tip of the iceberg. To compare your needs to this study, a more detailed account of the report can be downloaded here.


Author Stefanie Chernow: Stefanie Chernow is Content Manager at Valiantys. She is an international social media and digital strategy communications specialist. Adept at creating and developing organizational brands in the journalism, non-profit and IT sectors.