I’m looking into running a ScyllaDB cluster on AWS. How do I determine sizing?
Is it better to run a few large instances or many smaller ones?
Larger nodes are easier to manage and their resources aggregate much more performance and speed. Think of applying a patch or performing a rolling restart in a hundred small nodes, also the number of failures would be higher so you’d expect more often maintenance.
A disadvantage would be: since for AWS the storage size increases proportionately the larger the instance type, more powerful instances would become an expensive option in cases when the user doesn’t have a huge dataset. But still it may be recommended depending on the throughput and latency requirements.
A benchmark would help to find the sweet spot, as well as other details that can be found in a nicely explicative Webinar about this exact subject:
I recommend watching it!
1 Like