ServiceControl as an application can be used to process the entire message load of a system. This article provides general guidelines, recommendations, and performance benchmarks to help determine the resources to provide for a production environment. To identify the hardware specifications for any system environment, a combination of testing with the system and the information provided below will need to be used.
General recommendations
- Install ServiceControl on a dedicated server in production.
- 6 GB of RAM minimum
- 2 GHz quad core CPU or better
- Database path located on disks suitable for low latency write operations (fiber, solid state drives, raid 10), with a recommended IOPS of at least 7500.
Server performance monitoring
Due to changes in the system it supports, the requirements for a server hosting ServiceControl can change over time. It is highly recommended that monitoring of the CPU, RAM, disk I/O, and network I/O for the server running ServiceControl be included.
Real disk, CPU, RAM, and network performance can be monitored with the Windows Resource Monitor and/or Windows Performance counters.
Storage
It is recommended to:
- Store ServiceControl data on a dedicated disk. This makes low-level resource monitoring easy and ensures different applications are not competing for storage IOPS.
- Disable disk write caching to prevent data corruption if the (virtual) server or disk controler fails. This is a general best practice for databases.
Benchmark data
ServiceControl version 3.0.0 was tested to validate performance improvements made between version 2 and version 3.
Hardware used
- 16 vCPU
- 64 GB RAM
- 2x7500 IOPS striped disk dedicated for the database
Audit message processing throughput
Message Size | Messages Per Second |
---|---|
13 KB | 140 msgs/s |
66 KB | 80 msgs/s |
Disk usage
While disk usage was not captured across all tests, for a scenario using a 66 KB message size and storing 450,000 messages the total database size was 4 GB.
Suggestions to improve performance
More RAM
The embedded RavenDB will utilize additional RAM to improve indexing performance.
Message size / MaxBodySizeToStore
In general, the smaller the message, the quicker ServiceControl will be able to process audit records. Consider using smaller messages. For larger message payloads, consider using the data bus feature.
In addition, for audit messages, lower the ServiceControl/
setting to skip storage of larger audit messages. This setting will only reduce load if a non-binary serialization is used.
ServiceControl/MaxBodySizeToStore
limit.Use a dedicated disk for the database
Use a dedicated disk for the ServiceControl database path.
Additionally, it is possible to store the embedded database index files on a separate disk. Use the Raven/
setting to change the index storage location.
Azure disk limitations
Using multiple 7500 IOPS disks in striped mode in Azure may not improve performance due to increased latency; consider scaling out ServiceControl to multiple instances.
Scale out
If it is not possible to scale up ServiceControl to handle system volume, partition audit processing between multiple instances of ServiceControl. See Multiple ServiceControl Instances for more details.