Deployment sizing and capacity planning
The hardware capacity you need to plan for depends on your community's page view load.
Estimating deployment size
The usual calculation for estimating page views in preparation for a site deployment is to use the known page views per day and add 20% to the total as a buffer. (If you want to read more about the math behind this calculation, you can request the Jive 5 Performance Whitepaper from Support: it explains in detail how Jive is load-tested.) If you're replacing an existing site (such as a customer forum or intranet), you can use the page views from that site plus 20% to estimate the required resources. If your site is new, you can use page views from a comparable site. A page view is defined as a full page download: smaller interactions, such as AJAX actions, are non-significant for this calculation.
Note that the page view counts shown in the table below are based on averaging page views over a day. However, many sites have dramatic differences between peak usage and off-hour usage that are not adequately reflected in the daily average. If this is the case with your site, you should size for page views based on peak hours to avoid performance problems during high-traffic periods.
Required capacity
Deployment size | Required servers | CPUs per web application node | Memory per web application node |
---|---|---|---|
Small Fewer than 100,000 page views per day |
|
6 | 4GB heap, 10GB total |
Medium 100,000 to 500,000 page views per day |
|
6 or more | 6GB heap, 12GB total |
Large More than 500,000 page views per day* |
|
6 or more | 8GB heap, 16GB total |
* We typically use the following calculations to determine peak page views per second: 500,000 / 8 hours (for example, business hours) / 60 minutes (per hour) / 60 seconds (per minute) = approximately an average of 17 page views per second during business hours, assuming an even distribution of load across those 8 hours. Using this model, we expect to see spikes of up to 25-30 page views per second at peak.