Skip to Main Content
DoiT logo
Status Already exists
Pillar Optimization
Created by Guest
Created on Apr 28, 2021

Create a Full-Featured CUD Optimization Tool

The CUD charts in Cloud Analytics are a great tool I've been using to help determine how much a customer should consider committing to. I think there's a lot of opportunity to build upon this to automate a lot of the manual effort we do on top of this. Before I can ever recommend an amount to commit to, I always have to start with a conversation with the customer. However, there are several data points I have to gather as well that could be automated: - How consistent is the compute usage throughout the year? What's the lowest usage over the last 12 months? - How much does hourly compute fluctuate throughout the week and month? - What's the lowest hourly usage point this hits? That's the number we should consider committing to. - Determine this number and shows these graphs of usage over time for each instance type (N1, N2, E2, GPU). Determine for both CPU and RAM. Given those "recommended" commitment numbers, we should be able to show: 1) the amount to commit to in relevant units (cores, RAM) 2) the cost of those commitments, given either 1Y or 3Y CUDs 3) the savings associated with those CUDs 4) how those costs/savings fluctuate if I modify the amount (ie. if I commit to X% of 1Y and Z% of 3Y CUDs)
  • Attach files
  • Guest
    Reply
    |
    Apr 28, 2021
    If we want to really get advanced, we could integrate with Stackdriver Logging to capture resource utilization on each instance and determine which instances should be right-sized and auto-scaled.