Hello,
is there any documentation on the sizing we should consider when we want to deploy Renku on our own infrastructure? We are looking for a small to medium setup to begin with.
Thanks
Václav
Hello Vaclav,
unfortunately we do not have any documentation on the topic yet (although we are taking the point and plan something for the future!).
In our development Kubernetes clusters (that represents kind of the minimal deployment for Renku, monitoring, logging and other support services), we employ three nodes with 4 cores and 8GB of memory each. This is a comfortable enough base-line to have the essentials services running.
On top of that you will want to deploy a pool of nodes dedicated to your users, where the users-sessions can run, sizing this pool is straightforward as it directly relates to how many cores/memory you want your users to be able to access, multiplied the number of concurrent users you will be expecting. The nodes pool dedicated to users sessions is thus particularly well-suited to autoscaling as it can scale up or down depending on how many users are using Renku at a given time.
As for the storage we strongly recommend some form of Network Attached Storage, both for the services running on the cluster, as well for the scratch space attached to the users sessions, to ensure that the root disk of the nodes running the cluster does not get filled thus causing instability.
Let me know if you have further questions.
Cheers,
Alessandro.
Here are some current specs for Renku@UniFR which is mainly being used for teaching.
We have chosen to have only one of the master nodes that @aledegano mentioned (i.e. 4 cores and 8GB of memory). We also have two worker nodes with 8 vCPU and 32 GB RAM. In addition, we have 4 more nodes of this flavour dedicated to user jobs.
In all, I would imagine we have a small to medium setup and hope this helps you get started.