When sizing a VxRail cluster for implementation of our solution, it is important to first allocate the capacity that is needed to run any management VMs for the VMS and CV applications. The following management node types should be carefully designed for the proper scale and HA requirements of the deployment.
- Milestone XProtect
- Redundant SQL Server Databases
- Redundant Directory Servers (optional)
- Backup Archivers (optional)
- Redundant Web Services Consoles
- Redundant Databases
- Redundant Web consoles
- Redundant Databases
This design does not include recommendations for every VM and service that could add benefit from an HA implementation in a mission-critical environment. Our design should give a reasonable starting configuration for each type of VM. We note the sizing specifics for each application in the following sections.
Each VxRail node has the same configuration as described in the Physical architecture section.
For mission-critical safety and security solutions, it is necessary to have sufficient storage to archive video content for analyzing historical events. The retention period varies per customer and can be multiple PB in scale for large environments.
The technology used for VMS recording during this testing was the PowerScale platform. We have previously validated with the XProtect VMS and found that it provides flexible, scale-out file storage with the performance needed by large-scale camera networks. PowerScale is easily configured with XProtect for Tier 2 storage.
For more information about Dell Storage with Milestone, see Sizing Guide-Dell EMC storage with Milestone XProtect.
Cameras per node
Each VxRail node in the cluster is capable of hosting up to three Milestone XProtect primary recorders processing 70 camera streams each. We were also limited to a maximum of one Ipsoteck processing server and one BriefCam Alert Processing VM per node, each with one A40 GPU allocated.
- We tested with 840 camera simulator streams at an average of 3.2 Mb/s with 12 primary recorders.
- We tested processing 30 analytic camera streams with Ipsotek and 50 streams for BriefCam per node used for Face Recognition and Restricted Zone generating real-time alerts.
As shown in the Performance validation section we tested running 210 cameras across three recorders, the total server bandwidth consumed is less than 640 Mb/s.
Similarly, if processing for fewer analytics cameras is needed then it is possible to remove one or more Nvidia A40 GPUs from the VxRail node.