IT Unconference

Five Things a Sysadmin should do differently in Cloud

Proposed By
Richard Guo
Notes

  1. (Linux) Ideally, No more Swap

    1. The last thing anyone wants to do is troubleshoot performance problems

    2. (Preference) It is better for a component to die and be replaced, than to perform slowly

    3. Move toward: clustered and autoscaling architectures

  2. Use (and abuse) disks

    1. Configure as many disks as is reasonable

      1. Dedicated disk to take advantage of minimum IOPS guarantees

    2. Use disks to organize data

    3. Use disks to separate or combine IOPS performance

  3. Ditch partitions, use the whole disk

    1. Partitions are an obsolete construct. 

    2. Cloud disks are easily expandable.

    3. Snapshots provide easy recovery mechanism

    4. See Section 2 (Use and abuse disks)

  4. Take advantage of startup scripts

    1. AWS/GCP user-data/cloud-init

    2. Azure - Startup task

  5. Go ahead, make copies

    1. Copy that disk

    2. Copy that VM

    3. Spin up time: measured in minutes

Year
2019