[pve-devel] [PATCH perl-rs 1/1] pve-rs: resource_scheduling: allow granular usage changes

[Fiona Ebner]

Am 16.10.25 um 5:34 PM schrieb Daniel Kral:
> On Thu Oct 16, 2025 at 12:32 PM CEST, Fiona Ebner wrote:
>> Am 30.09.25 um 4:21 PM schrieb Daniel Kral:
>>> Implements a simple bidirectional map to track which service usages have
>>> been added to nodes, so that these can be removed later individually.
>>>
>>> The StaticServiceUsage is added to the HashMap<> in StaticNodeInfo to
>>> reduce unnecessary indirection when summing these values in
>>> score_nodes_to_start_service(...).
>>>
>>> Signed-off-by: Daniel Kral <d.kral at proxmox.com>
>>
>> Reviewed-by: Fiona Ebner <f.ebner at proxmox.com>
>>
>>> ---
>>> I started out adding and removing the service_usage in StaticNodeUsage
>>> on each add_service_usage_to_node(...) and remove_service_usage(...)
>>> call individually, but I think summing those up every time is better for
>>> numerical stability..
>>
>> This can go into the commit message too if worded a bit less informally ;)
> 
> Will do! Even though I asked myself if the concern about numerical
> stability really holds true here if we don't go for ha #9 as then these
> values will only change over a single manage(...) run and I'd guess that
> most users would use floating-point values which can be represented
> exactly (0.0, 1.0, 2.0, 3.0, ... and 0.5, 0.25, 0.125, ...). For these,
> numerical stability shouldn't be a problem at all, or not?

I guess cpulimit for containers is the obvious exception ;)