Kubernetes a black hole of unpredictable spend, according to new report
- Reference: 1624984093
- News link: https://www.theregister.co.uk/2021/06/29/kubernetes_spend_report/
- Source link:
The new report is based on a survey of the CNCF and Finops Foundation communities and although it had only 195 respondents, there was a "strong enterprise representation".
Spending on Kubernetes was up, according to the survey, with 67 per cent reporting an increase of 20 per cent or more over the last 12 months, and 10 per cent spending more than $1m per month on their deployments.
[1]
Kubernetes, as Reg readers know, is a means of orchestrating containerised applications, so its actual cost is mainly that of the resources it consumes: over 80 per cent compute (such as VMs on AWS, GCP or Azure) and a bit on memory, storage, networking and so on.
What's the Finops Foundation? The Finops Foundation was set up in 2019 by Cloudability, a cloud financial management company now part of Aptio.
In 2020 it became part of the Linux Foundation, alongside Kubernetes overseer Cloud Native Computing Foundation, and aims to "advance the discipline of cloud financial management." Sponsors include Aptio, Atlassian, VMware and Google Cloud.
Basically, it is a kind of alliance of those hoping to manage and reduce their cloud spend, including Nationwide, Spotify and Sainbury's, with companies that want to sell you cost management solutions, though the Foundation [2]has promised to focus on "best practices beyond vendor tooling."
The cost of such things as training IT staff and maintaining complex deployments was sadly not a focus of the report. Rather, its key point was that many organisations have only a loose grip on what they are spending.
"The vast majority of respondents… either do not monitor Kubernetes spending at all (24 per cent), or they rely on monthly estimates (44 per cent)," the report said.
[3]
[4]
Of those which do use tools, the favourites specified were largely the cloud provider's tools such as AWS Cost Explorer, Azure Cost Management, and GCP Cost Management, though there was a significant showing for the open source [5]KubeCost with 12 per cent adoption. The humble spreadsheet came in at 5 per cent.
Accurate prediction of Kubernetes costs is a challenge
Less than 25 per cent of those surveyed said they could accurately predict how much they’d spend on Kubernetes to within 5 per cent of actual cost. However, the figure climbed to circa 60 per cent when they were asked if they could predict spend to within 10 per cent of the real cost.
Over 20 per cent apparently stated "we do not predict our cloud bill" though that was not correlated to size of organisation so may represent mainly small deployments.
There is evidence though that the exact size of the monthly Kubernetes bill comes as a bit of a surprise to many companies, which is far from ideal given the high and increasing cost of the technology.
[6]What the heck is FinOps? It's controlling cloud spend – and new report says it ain't easy
[7]Enterprise databases deployed in Kubernetes? Proceed with caution, warns seasoned analyst
[8]What Microsoft's Windows 11 will probably look like
[9]VC's paper claims cost of cloud is twice as much as running on-premises. Let's have a look at that
According to Finops Foundation, the solution is to do a better job of tracking expenditure. Suggestions include combining real-time Kubernetes cluster costs with additional costs such as database instances and cloud storage use, drilling down to the nodes, pods and persistent volumes and carefully tagging and allocating each resource, and to automate cost reports to track trends.
Controlling and reducing Kubernetes costs is not inherently different from managing other kinds of IT costs, but the abstraction it offers perhaps makes it easier to miss the detail of the resources consumed. There is also the challenge of scaling appropriately to the demand, with over-provisioning all too easy.
[10]
"Worship at the altar of 'turn that shit off'. You're charged for what you forget to turn off," cloud cost consultant Corey Quinn [11]told us last year. The fact that it is Kubernetes should be no exception.®
Get our [12]Tech Resources
[1] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_offprem/front&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=2&c=2YNuYAQL0Wx83s-M3S-D@3AAAAFg&t=ct%3Dns%26unitnum%3D2%26raptor%3Dcondor%26pos%3Dtop%26test%3D0
[2] https://www.finops.org/about/
[3] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_offprem/front&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=4&c=44YNuYAQL0Wx83s-M3S-D@3AAAAFg&t=ct%3Dns%26unitnum%3D4%26raptor%3Dfalcon%26pos%3Dmid%26test%3D0
[4] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_offprem/front&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=3&c=33YNuYAQL0Wx83s-M3S-D@3AAAAFg&t=ct%3Dns%26unitnum%3D3%26raptor%3Deagle%26pos%3Dmid%26test%3D0
[5] https://www.kubecost.com/
[6] https://www.theregister.com/2021/02/10/finops_controlling_cloud_spend/
[7] https://www.theregister.com/2021/06/24/postgres_kubernetes_danger/
[8] https://www.theregister.com/2021/06/15/windows_11_leak_microsoft/
[9] https://www.theregister.com/2021/06/02/andressen_horowitz_paper/
[10] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_offprem/front&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=4&c=44YNuYAQL0Wx83s-M3S-D@3AAAAFg&t=ct%3Dns%26unitnum%3D4%26raptor%3Dfalcon%26pos%3Dmid%26test%3D0
[11] https://www.theregister.com/2020/09/03/cloud_control_costs/
[12] https://whitepapers.theregister.com/
Re: Green graphs
It's got labels on the side and the bottom.
Contended for worst graph of the year
This doesn't seem to be about Kubernetes spend at all
As the article mentions, it's really (mostly) node cost, storage cost and network cost. How you run your software on that infrastructure, and what limits you put in place, will be the same regardless of if you run it all manually, have an orchestrator, or anything in between.
This might be more appropriately a "cloud spend" tracker, as it's certainly easy to run without limits in the cloud. If you manage your own hardware/VM estate you're probably keeping a closer eye on your infrastructure growth.
"Worship at the altar of 'turn that shit off'.
With most businesses just wanting new features and thinking it is more cost effective to spend money on more CPU / RAM / Storage / Network than developers spending time optimizing the code is there any wonder that most places don't care about the cost of running the code.
I've been in ALOT of places where the infra costs are a problem that the Ops team get yelled at, not the developers. Most developers I've worked with have never profiled their apps to give realistic cpu or memory requirements, that's something that QA should find out with load testing right ?
Then add to the mix how many places are selling their code for others to run, what incentive do they have to optimize?
WHo has time to optimize? The name of the game is to slap as many features together as possible as fast as possible and ship it!
"slap as many features together as possible as fast as possible and ship it!"
And don't waste time and effort testing. The customers will do that for you. Much faster and more cost effective overall than waiting for a test team to dink around with the product for days or weeks. If it compiles clean and links, it's good to go.
Maximize efficiency
"slap as many features together as possible as fast as possible and ship it!"
And don't waste time and effort testing. The customers will do that for you. Much faster and more cost effective overall than waiting for a test team to dink around with the product for days or weeks and then having the customers find a whole lot of additional problems as they always do.
If it compiles clean and links, it's good to go.
That's precisely the point and what the report fails to address. It's the elephant in the room
This is why DevOps is a good idea. Build and run it.
Marketing press release
I'm seriously trying to understand what is here. As mentioned, this is a lot about cloud spending, and almost negligently about K8s spending. Beyond that, this survey was obviously designed by someone with no understanding of the underlying issues.
60% of companies can predict spending to within 10%. What percentage can predict usage to within 10%? If your spending prediction is significantly more accurate than your usage prediction, then you are wasting capacity. Full stop. And yet this article speaks as if the goal is to accurately predict spending.
I can make it really easy to predict the costs. Build a datacenter, and fill it with 2x as much capacity as you need for your business. Donate all your idle samples to protein folding. Done.
The goal is to minimize costs while delivering an acceptable level of service. Predictability of costs is in direct conflict once the basics are covered. In particular, seizing a sudden opportunity to "turn that s*** off" results in an unpredicted cost of 0 for that resource.
Yawn!
Nice to see accurate comparison of Kubernetes' ability to meet expectations with that of other container/cloud/VM architectures. There's nothing like a solid baseline.*
Just checked my calender and this month is named Kubernetes-Bashuary.
* All together now, "And this has nothing like a solid baseline."
Green graphs
Odd to see a bar graph with a breakout that is all in the same color o.O