David hernandez

@david2127
6 Followers
17 Following
30 Posts

Anyone has implemented gang scheduling for Ray clusters with per ray cluster granularity. I do not need ”fancy” borrowing or fairness. But I do want per Ray cluster not per let’s say namespace shared quota. Any tips?

#Ray #Kueue #Volcano

Anyone has implemented gang scheduling for Ray clusters with per ray cluster granularity. I do not need ”fancy” borrowing or fairness. But I do want per Ray cluster not per let’s say namespace shared quota. Any tips?

#Ray #Kueue #Volcano

Anyone has experimented with spot VMs for celery workloads and set task_reject_on_worker_lost = ”True”?

#celery #python

It is possible to use cloud autoscaler alongside karpenter on the same cluster?

#k8s #karpenter

Has anyone tried successfully the bpfman https://bpfman.io/main/ quick start on kind cluster where cilium was installed

I am not 100% sure but I think while trying to mount bpf fs it messes up with cilium bpf fs

bpfman

Bandwidth was, as expected higher for the SSD premium V1 without NVMe controller. But a colleague pointed out about Disk burst feature on azure plus the max uncached burst support for the VM size and it matches the results
Have not checked the bandwidth results. First task tomorrow morning
Have tried also SSD premium V2 logical sector size 4k NVMe controller and the results are 3 K IOPS tried also same SSD premium V1 again (logical sector size is 512 B) and got only 14 K latency for the NVMe experiments are also higher than the current setup. I know I might not be testing the current workload applied by mongo. But am I missing something about SSD performance in azure managed disks #azure #performance #mongodb
Looking into improving performance of our mongodb self hosted in azure. Focusing only on the disk for now as according to gravaba dashboard hits the 100% utilisation during heavy reads periods. Currently is running on Premium SSD 2T V1 (7500 IOPS) with caching set to ReadOnly, benchmarks by using fío results in 23 K IOPS and ~5 ms latency #azure #performance #mongodb
Current code in pandas, finds columns of type list and apply a UDF to deal with single quotes inside double quotes, to do that it cast the whole column as a string “[]” and then apply the function row by row. How would you go about doing this using polars #pandas #polars #python