Spot
The pricing of Spot is really too tempting. As a student with little money, this price seems very cost-effective, but it is taken up too quickly. What are you guys doing with Spot? Sometimes this thing It can only be used for 5 minutes.
3 Replies
use cases for spot instances:
Spot Instances are spare EC2 capacity that AWS offers at a discounted price compared to On-Demand instances. However, they can be interrupted with short notice when AWS needs the capacity back. This makes them suitable for certain use cases but not for others.
Some common use cases for Spot Instances include:
Batch processing jobs
Data analysis and scientific computing
CI/CD and testing environments
Rendering and transcoding
Big data processing with frameworks like Hadoop or Spark
Machine learning and AI training
The short availability you're experiencing (only 5 minutes sometimes) is likely due to high demand for the instance type you're trying to use. To mitigate this, you could:
Try different instance types( other gpus )
Use Spot Fleet to request instances across multiple instance types and Availability Zones
Be flexible with your timing, as prices and availability fluctuate
Remember, Spot Instances are not suitable for applications that require consistent, uninterrupted compute capacity. For critical or production workloads, On-Demand or Reserved Instances are usually more appropriate. ( aws )
So its alike
the use cases are same, but you gotta use some pilot/orchestrator tools like skypilot to make sure it runs smoothly, or else it will be a hard time managing spot jobs
what are you using runpod for then?
In fact, I haven't started to do any work with spot, I just use the on-demand GPU to train some models. I noticed the price of Spot, but I didn't expect that it is not suitable for training models at all.
Ah ya as I said, I guess you'll have to use some kind of orchestrator to do that