Compute and memory optimized flavors

ovh / public-cloud-roadmap

Agile roadmap for OVHcloud Public Cloud services. Discover the features our product teams are working on, comment and influence our backlog.

https://www.ovhcloud.com/en/public-cloud/

184 stars 5 forks source link

Compute and memory optimized flavors #456

Open matmicro opened 10 months ago

matmicro commented 10 months ago

As a Managed Database user, I need to be able to choose a specific kind of instance (CPU oriented, RAM oriented, or Generic).

For example in my case, i am running a DB1-15 because i need 4 cpu cores. But my database is using around 20% of RAM.

Thus in order to reduce cost, i would be able to run a database flavor with 4 CPU but less RAM. So as for PCi instances, it would be useful to be able to choose the instance flavor types, or have more choices in D flavor lists.

thild42 commented 10 months ago

Thank you for the suggestion, I understand the need.

I do need to check internally if providing flavors with optimized CPU, RAM, storage or even aligned with used cases like basic, general or optimized would make sense.

I keep this issue open to share progress.

matmicro commented 8 months ago

Any progress on this ?

thild42 commented 8 months ago

We still are considering the different aspect of this feature request. Currently we identify a first need to leverage memory optimized instance for Redis.

matmicro commented 8 months ago

I anderstand the RAM aspect for Redis.

You should also consider CPU usage for Postgres. It is currently a big bottleneck for me that make me think about moving out of Managed DB as ratio cost/perf is really expensive in regards of poor perf.

ps: CPU bottleneck may be linked also with network disk perf

thild42 commented 5 months ago

Hey, we are working on compute and memory optimized flavors for managed databases.

Assuming we build these flavors on third generation of public cloud instances, you would be presented with 3 choices if you want to upscale your flavor resources:

(1) Upscale the flavor: double the vCores, the memory, the base storage and the price.
(2) Upscale compute: double the vCores only, memory and base storage stay the same. Save 22% compared to a flavor upscale.
(3) Upscale memory: double the memory only, vCores base storage stay the same. Save 61% compared to a flavor upscale.

Please tell what you think about it.

Note: prices and saving are based on third generation of public cloud instance hourly pricing without savings plans. These prices and savings are rough estimation, not definitive, and need to be refined. I share them as order of magnitude to discuss.

matmicro commented 5 months ago

Ok so if i well understand the cases, (1) is the actual behavior. And (2) and (3) will be added by proposing more flavors (2 = C3 and 3 = R3).

This looks perfect ! Any idea about ETA for this to be released ?

Thanks !

thild42 commented 5 months ago

Yes that's a good summary. No ETA yet as I am just beginning to refine this topic.