Confluence 5.6 has reached end of life
Check out the [latest version] of the documentation
Server administrators can use this guide in combination with the free Confluence trial period to evaluate their server hardware requirements. Because server load is difficult to predict, live testing is the best way to determine what hardware a Confluence instance will require in production.
Peak visitors are the maximum number of browsers simultaneously making requests to access or update pages in Confluence. Visitors are counted from their first page request until the connection is closed and if public access is enabled, this includes internet visitors as well as logged in users. Storage requirements will vary depending on how many pages and attachments you wish to store inside Confluence.
The values below refer to the minimum available hardware required to run Confluence only, eg the minimum heap size to allocate to Confluence is 512mb. You will need additional physical hardware, of at least the minimum amount required by your Operating System, and any other applications that run on the server. Also please note that these are a guide only, and your configuration may require more.
On small instances, server load is primarily driven by peak visitors.
5 Concurrent Users
25 Concurrent Users
Note: Please be aware that while some of our customers run Confluence on SPARC-based hardware, Atlassian only officially supports Confluence running on x86 hardware and 64-bit derivatives of x86 hardware. Confluence typically will not perform well in a tightly constrained, shared environment - examples include an AWS micro.t1 instance. Please be careful to ensure that your choice of hosting platform is capable of supplying sustained processing and memory capacity for the server, particularly the processing-intense startup process.
These are example hardware specifications for non-clustered Confluence instances. It is not recorded whether the amount of RAM refers to either the total server memory or memory allocated to the JVM, while blank settings indicate that the information was not provided
Accounts | Spaces | Pages | CPUs | CPU (GHz) | RAM (MB) | Notes |
|---|---|---|---|---|---|---|
150 | 30 | 1,000 | 1 | 2.6 | 1,024 |
|
350 | 100 | 15,000 | 2 | 2.8 | 1,536 |
|
5,000 | 500 |
| 4 | 3 | 2,048 |
|
10,000 | 350 | 16,000 | 2 | 3.8 | 2,048 |
|
10,000 | 60 | 3,500 | 2 | 3.6 | 4,096 |
|
21,000 | 950 |
| 2 | 3.6 | 4,096 |
|
85,000 | 100 | 12,500 | 4 | 2.6 | 4,096 | 3 machines total: application server, database server, Apache HTTPD + LDAP tunnel server. See Accenture's slides and video for full details (That link isn't working, but the slides can be found here) |
When planning server hardware requirements for your Confluence deployment, you will need to estimate the server scalability based on peak visitors, the editor to viewer ratio and total content.
Confluence scales best with a steady flow of visitors rather than defined peak visitor times, few editors and few spaces. Users should also take into account:
These values are largest customer instances reported to Atlassian or used for performance testing. Clustering, database tuning and other performance tuning is recommended for instances exceeding these values.
Most Spaces | 1700 |
Most Internal Users | 15K |
Most LDAP Users | 100K |
Most Pages | 80K |
All page content is stored in the database, while attachments use either the database or file system. For example, the wiki instance you are reading now uses approximately 2.8 GB of database space and 116 GB of disk space. The more attachments you have, the more disk space you will require.
Private and public comparison
Private instances manage their users either internally or through a user repository such as LDAP, while online instances have public signup enabled and must handle the additional load of anonymous internet visitors. Please keep in mind that these are examples only, not recommendations:
Use Case | Spaces | User | Editors | Editor To | Pages | Page Revisions | Attachments | Comments | Total Data | Notes |
|---|---|---|---|---|---|---|---|---|---|---|
140 | 11,500 | 1,000 | 9% | 8,800 | 65,000 | 7,300 | 11,500 | 10.4 |
| |
130 | 180 | 140 | 78% | 8,000 | 84,000 | 3,800 | 500 | 4.5 |
| |
100 | 85,000 | 1,000+ | 1%+ | 12,500 | 120,000 | 15,000 |
|
| Accenture - see slides and video for full details (That link isn't working, but the slides can be found here.) |
For large instances, it may be worthwhile contacting an Atlassian Expert for expertise on hardware sizing, testing and performance tuning. Simply contact a local Expert directly or email our Experts team for a recommendation.
Here is a breakdown of the disk usage and memory requirements for this wiki, as at April 2013:
Database size | 2827 MB |
|---|---|
Home directory size | 116 GB |
| Average memory in use | 1.9 GB |
Data | Relevant Table | Rows | Size |
|---|---|---|---|
Attachment metadata | attachments | 193903 | 60 MB |
Content and user properties | os_propertyentry (?) | 639737 | 255 MB |
Content bodies (incl. all versions of blogs, pages and comments) | bodycontent | 517520 | 1354 MB |
Content metadata (incl. title, author) | content | 623155 | 459 MB |
Labels | label (5982, 1264 kB), content_label (134151, 46 MB) | 140133 | 47.2 MB |
Users | users | 38766 | 6200 kB |
Data | Files | Size |
|---|---|---|
Attachments (incl. all versions) | 207659 | 105 GB |
Did-you-mean search index | 10 | 14 MB |
Office Connector cache | 3506 | 456 MB |
Plugin files | 1851 | 669 MB |
Search index | 448 | 3.9 GB |
Temporary files | 14232 | 5 GB |
Thumbnails | 86516 | 1.7 GB |
Usage index (now disabled) | 239 | 2.6 GB |
Note: not all files are shown, and average file size may vary between instances.