N2W Software

            Recommended instance sizes/volume types for CPM Server instances

            Here are the recommended instance sizes for CPM Server instances:

            Up to 100 instances    - t2.small
            Up to 200 instances   - t2.medium
            Up to 500 instances   - t2.large
            Up to 1000 instances  - t2.xlarge/m4.xlarge
            Up to 2000 instances - t2.2xlarge/m4.2xlarge

            This recommendation is based on the following assumptions:
            1) Instances have 3 volumes on average.
            2) No more than 20% of instances have Agents performing VSS/scripts.
            3) No more than 30% of instances are being backed up simultaneously (within 30 minutes from each other).

            Exceeding these assumptions may require larger instance sizes.

            Important - for 200 and more protected instances (or when running backups more frequently than hourly) we recommend to switch the CPMDATA volume's type from "gp2" to "io1" with minimum 1000 IOPS:

            You can switch the volume's type with zero downtime by following this guide: http://docs.aws.amazon.com/AWSEC2/latest/UserGuide/console-modify.html
            Since achieving 1000 or more IOPS requires resizing the volume, please follow this guide to allow OS to see the new size correctly: https://support.n2ws.com/portal/kb/articles/resizing-the-cpmdata-volume

            Using "gp2" volume for 200+ instances (or frequent backups) may cause severe performance issues, including failing backups/DR and unexpected restarts of CPM Agent/Server.
            For best results, it's recommended to switch to the M4 family of instances, which is EBS-optimized and can take full advantage of the io1 volume.

            For over 1000 instances or 3000 volumes, you may need to raise IOPS further.
            Please note, that CPM is more IO-intensive than CPU-intensive, improving IO performance will usually gain better results than improving CPU performance.

            When using T2 instances, it is recommended to monitor the CPU Credit Balance to make sure it doesn’t reach 0, as it may affect the instance’s performance.
            Please read the following KB to better understand how CPU credit balance is calculated and how draining it may affect your instance’s performance.

            When using GP2 volumes, it is recommended to monitor the BurstBalance to make sure that the volume can provide more than the default provisioned 100 IOPS: http://docs.aws.amazon.com/AmazonCloudWatch/latest/monitoring/ebs-metricscollected.html


            Help us to make this article better
            0 0