Hi all, I'm helping a colleague create a charter f...
# general
r
Hi all, I'm helping a colleague create a charter for a new FinOps team within our company. Would love to hear from you the features/requirements your FinOps teams have leveraged to maximize efforts for cost savings/optimization and tco.
p
I cannot express enough how important it is to find a way to make individual teams responsible for their own costs (depending on your scale, ideally cloud costs come out of team budget). A team which can spend without constraint will design without constraint.. never ends well! 😂
m
A couple things - 1. Make the costs visible - A comment on a PR with the new estimated cost is great, otherwise budget and cost trend alerts - Reports that require manual review are very ineffective 2. Attribute costs based on ability to fix, not on the payer e.g. If a security team require specific audit logs and control the log settings the cost should be allocated to them, if a team can change those log settings themselves - then they should have the cost - This requires multiple attribution models 3. Provide a total cost of ownership (including discounts, reservations, licensees, subscription usage and inter-team cross charges) 4. Try and attribute cost to individual products / profit centres / customers - Cost centres are often easy to reuse and abuse 5. Consider the distinction between R&D and cost of sales, 6. Enable cross-charging of shared resources and make it visible 7. Might be controversial, but include labour cost - Its often easy to recommend a change based on the cash cost, but overlook the labor cost 8. If possible build financial ratios based on quality, labour, security etc. 9. Combine FinOps and a teams normal backlog e.g. if you have automation open and close issues in a teams backlog - but be aware of crying wolf and false attribution
j
Its surprisingly hard to attribute costs to teams. If anyone has good solutions I'd love to know what they are.
m
Tagging of resources is the most common - You enforce a tag on all resources with a Cost Center / Owner / Team / Product and automatically delete (new) resources without those tags
j
you need to be mindful about specific platform components that may make attribution harder, i.e. kubernetes namespaces for shared / multi-tenant deployments
the other meta topic is more around "how do you attribute cost for a given platform team?" do you attribute the cost to the platform team enabling a capability or to a downstream consumer using that capability? you can apply this frame of thinking to both compute/memory and data/storage platform components
it's an organization and a platform team specific decision that needs to be made. there is no right or wrong on approach, you just need to advocate for a philosophical choice but apply it consistently across all instances in your environment
m
I think shared resources should be billed back to the consumer (unless doing so drives the wrong behaviour like opting out of security) - there are many strategies for this e.g. split cost evenly, by platform usage, by consumption (e.g. per GB of log) - tooling though is not great in this area.
j
nope, it really isn't and it would vary quite a bit based on each organization and approach. unless a given component's cost footprint is relatively "small" (as defined by the organization's budgets) in terms of real dollars, most participants wouldn't really be open to the idea of "split costs evenly"
m
In large orgs with multiple business units / companies I think you will find it happens more often than not, even if just at a an accounting level - e.g. the cost of the central IT is shared across all divisions / business units evenly or based on revenue or % of attributable spend. In order to get away from it you need a higher level of maturity.
j
Right, thats the simplistic model and no one is happy with it 🙂