Category: Practitioner

Do you want to know what is ahead for FinOps in 2023? This is Kostner’s take on what is ahead in the world of those managing costs of a major cloud platform like Azure, AWS and GCP.

Prefer video over text – scroll to the bottom of this page to see the full video.

The 3 major predictions are:

1. Cloud Cost Management on the CIO agenda - as well as the CFO

Cloud spend on major platforms like Azure, AWS, or GCP will eclipse all other IT expenditures to become the biggest cost item in IT. This will attract the attention of finance, in a way you likely haven’t been accustomed to. If IT can’t give credible answers to how it intends to manage and reduce what is to date on average 50% YoY growth, Finance is likely to take control – and that’s rarely a fun experience for IT.

2. Strategic Management of Commitments

Trading Agility for Affordability. I often joke that agility in cloud is 20% technology and 80% total lack of process. Well, going forward, expect to see more process. The hope is we can use a combination of the flexibility of cloud to allow after-the-fact cleanup, and the predictability and lifecycle of systems, to avoid the dreaded pre-approval processes. With the recent changes in policy, we need to be able to look at least 1 year into the future and make decisions about the stable portion of our environment. If we can’t, we’ll be missing out on at least half of savings.

3. Review, Spend and Clean-up

About half of all the attainable savings come from regularly reviewing the spend on installed environments with those who know that environment best. We regularly see identified savings of between 20 and 30% the first time this kind of spend review is performed, and making this a core process is a great way to keep that spend under control.

Next step?

So when your CFO realizes that your cloud spend has gone up a lot and it is looking like it is going to continue – having control over these two elements, presenting them, is a good way to make sure you get to keep as much agility and flexibility as possible – and still run as cost-efficiently as you possibly can.

Want more FinOps knowledge

Have your company decided to start looking into showback/chargeback for your public cloud costs? And have you been given the responsibility to set it up? This article gives you the 3 elements of showback/chargeback as well as the process of implementing it.

The whole point of chargeback is making the different business owners accountable for their cloud spend. Implementing a simple model first and then refining it over time will create understanding and ownership. This article gives you an overview so you can plan when to implement which steps – and not create obstacles early on for natural extensions you will want later.

To implement showback/chargeback in practice there are three elements you always need to go through.

Element 1: Who is spending?

For chargeback/showback you need to know exactly who uses what resources and who drives what costs in Azure. Keeping track of the different departments usage and checking their Azure bills is time consuming work. Tagging resources is the common approach, but in order to minimize administration we recommend a combination of the following:

Resource tagging: All resources with a cost center tag are allocated after these tags.

Virtual tagging: Enrollment accounts, subscriptions (or resource groups) are given a cost center tag, that will be used as tag, if there isn’t a tag on the individual resource.

Cost center for unmapped costs

We have seen other rules applied, but the above mentioned consistently deliver robustness and limited time spent administering allocations. You may want to start by allocating to only one or a few major internal customers and then continue with a wider group.

Even if you have decided on a strategy to rely on resource tagging make sure you include the other two levels. Not all costs and resources can be tagged, and you will never be 100% up to speed with resource level tagging.

Also, make sure that you have a process in place to clean up unmapped cost items monthly by assigning them a resource or virtual tag and a process to regularly check for validity of used tags.

Element 2: How much are they spending?

When your tagging is in place you have a clear picture of who owns which resources and the associated costs. But this is only the first element of chargeback – the second is to figure out how much they should be charged.

a) What rate to use?

Out of box all the clouds offer two different rates you could use for chargeback: a) actual costs, and b) amortized costs.

The actual cost is what reconciles directly with your monthly cloud bill. This means that if you prepaid something – the actual cost is zero. If you bought a reservation there will be a cost for the reservation, but the cost for the resource it is applied to will be zero.

Amortized costs are the cost where any committed use discounts are applied (depending on your cloud: reserved instances, compute savings plans, CUDs etc.). This gives a better representation of the real cost – but it comes with a flaw: if you manage commitments centrally, you cannot control who benefits from the commitments. Hence, a cost center might use the same quantity of resources in two subsequent months, but their charge may vary – not by a lot, but enough to cause confusion and extra work.

Blended rate – which is currently only an option in the AWS portal – ensures all that all resources covered by the same commitment plans are charged the same unit costs. This reduces variances from month to month. Your FinOps partner might – just like Kostner – offer a calculated blended rate if this is your preferred choice.

A few more options exist, i.e., standard cost, but we strongly recommend that you settle on either amortized cost or blended rate.

This is the one decision which is critical to get right when starting out – using the wrong rate might cause issues later – and changing the rate along the way will always create internal issues and lack of ownership.

b) Need to charge any additional costs?

Once you have settled on which rate you will be using the next decision is if you need to charge any additional costs, e.g.:

Add-on costs

In this step you decide which add-on costs to add that are driven by your cloud usage, but not invoiced through the cloud bill. This can be direct costs (i.e., costs for Microsoft licenses used for AHB), people costs enabling cloud development (CCoE, Cloud Foundation Teams, FinOps teams etc.), and more depending on your setup.

Costs of shared resources

On your cloud bill you may find costs that are shared between different cost centers e.g., Kubernets, micro services etc. You want to allocate these to the cost centers driving the costs – though not necessarily down to the last eurocent.

Shared expenses

Lastly, you will have items on your cloud bill which are not directly shared resources, but are nevertheless costs indirectly related to running the cost centers, e.g., shared infrastructure, nonutilized reservations, resources without tag etc.

When you first start out, these items may not appear very important – in fact we see many organizations where the cloud adoption is initially so business driven that there is not much shared cost personnel to consider allocating costs for.

However, it will become relevant so make sure that you prepare for this – at least in communication – right from the start.

Element 3: Practicalities

Coordinate with Finance

Make sure you involve Finance early on. You might find that there are existing policies that could guide you on the decisions above – and they will certainly have views on additional costs whether they should be charged by you or in a separate process.

Also, you want to ensure that the process and communication is aligned with your budgeting process – otherwise the line-item cloud costs will not be taken seriously.

Lastly, you want to ensure that you have agreed on the practical process with Finance:

When do they need a total cost (estimate) for the month?

When do they need the chargeback file?

Who and how do you reconcile between the bill, the estimated total, and the chargeback file?

Implementation

Before implementing chargeback, you need to be completely sure what it is you want, so you can explain to the organization’s department what it is they are being charged for. As every big change needs an adjustment time, it is a good idea to start implementing showback first, and getting the organization used to the idea that they have to keep an eye on their cloud spend. People need to see and hear things before they accept it, hence make them used to realizing the size of their own cloud spend before starting to charge them for it.

Supporting chargeback

Once implemented you have three tasks that are part of supporting chargeback operationally:

Providing the data to the cost center owners
Supporting the cost center owners with input on what they can do to control costs
Providing Finance with the necessary data

On 1. you not only need to provide your cost center owners with data – but also a meaningful representation of what they are spending on.

Also, you want to make sure that it is clear to the cost center owners that a number of savings options have already been handled centrally, i.e., by providing a “Pay-as-you-Go” cost and the costs they are charged. This will help build credibility around the FinOps efforts you are orchestrating.

To support the cost center owners, you need to understand their data about what is driving costs – and what initiatives can be started to reduce costs. Most likely you have centralized price optimizations (i.e., commitment documents), but you can support them on discussing major cost items, and cost architecture.

How to get started?

If you want to fasttrack your showback/ chargeback project and avoid mistakes others are doing, then make sure you consult with a FinOps partner.

Do you want guidance to decide on policies? Get an “out-of-box” PowerBI solution? Design your tagging for minimum administration.

Kostner’s FinOps-as-a-Service will give you all of this and much more – we would love to be part of your virtual FinOps team. Book an inspiration meeting with us today to get started on your chargeback.

Have you started hearing the word FinOps often? Or trying to wrap your head around how to manage your growing cloud costs? Without adding another time-consuming task to your team or removing the flexibility and agility of cloud in your organization?

This article gives you an overview of the why and how of FinOps so you can plan your FinOps activities and get started. In the bottom you’ll find the most important advice we’ll leave you with. Completely unintentionally… – you need to scroll through the whole blogpost to see it.

Why do you need FinOps?

The reason why you need to do FinOps is that you can save 50% of your bill through structured financial governance compared to an ungoverned cloud environment.

The savings come from 3 different sources:

Buying services as cheaply as possible and leverage all discount options
Only pay for what you need
Use the most cost-efficient services that fits your needs

FinOps is like cost management when you build a house

Just as when you are building a house – when you buy your materials, you want to buy the materials as cheaply as possible, and you might research where you can get them at discount. Second, you need to manage the quantity and make sure you do not buy more than needed so you later have to return them. Third, you look at the overall architecture and make sure you avoid using unnecessary expensive building blocks.

The new way to do financial management

Cloud calls for a new way to do financial management. The new thing is that we buy IT as a service and not large hardware and software investments. This gives flexibility and agility – but how do you ensure sound financial governance without creating a bureaucratic overhead sacrificing flexibility and agility – and adding a huge job on your scarce cloud resources?

Below we will go through the 4 steps of FinOps:

Visibility – know your costs

Create awareness of cloud costs by looking at your bill every month.

Visibility is the preliminary step to get your FinOps efforts going. If you’re not doing any FinOps initiatives yet, just taking the step of looking at your bill every month will put you in the mindset to start asking the right questions such as:

How much did we grow since last year?

What areas are driving the increase in costs?

Are we using everything we are paying for?

You need to know these things and track the development over time to avoid being blindsided by rising cloud costs. For example, a 5% increase one month may not seem like a lot, but if it continues for 6 or 12 months it’s completely different numbers.

Price optimizations – Buying services as cheaply as possible

Avoid overpaying for your services and leverage all discount options. This will save you 20-25%.

You can optimize price in many ways, but fundamentally it boils down to optimizing price, by working through the cloud service catalog, price, and discount models.

You might know the 3 essential ways to optimize on price – reservations /commitments, Azure Hybrid Benefit, and shared resources. But what makes price optimizations complex is the myriad of options there is within each of these, all affected by complex details, the state of your environment and the fact that Microsoft keeps changing the options for price optimizations. It’s not a difficult task – it’s just really time consuming and you need to ask yourself:

Does it make sense for you and your team to trawl through your whole cloud environment as well as your portal? Or would it be easier to get someone else to do it?

A tool is a great idea to do this as it is an optimal task for technology. With a tool your team doesn’t have to spend unnecessary time doing it.

While there is a tool side to price optimizations, i.e., using the organization tools in your portal, the main reason for not getting price optimizations done is uncertainty. What if our environment changes? What if I will move to a different service shortly? What if…?

The solution is investing in both tool and training. The tool makes you able to find the needle in the haystack and training gives you the necessary expertise to remove uncertainty and act on the recommendations from the tool.

Read more about price optimizations:

3 ways to get the most Azure for your money – and how to get it done!

3 indicators of a cost-efficient cloud

The 4 things you want to know as a CIO about cloud costs

Quantity management – only pay for what you need

Avoid paying for something you don’t use. This will save you another 20-25%.

In this step you need to get the different business owners involved in making decisions about shutting down services (lifecycle management), shutting services on/off, and buying the right capacity services (rightsizing).

To get the business units committed you need showback / chargeback to gain an overview, of which department is using what. Also, adding chargeback, incentivizes the different departments to take ownership and manage their cloud costs in a more structured way if they know they will gain the economic benefit of their own efforts.

Showback / chargeback is a moving target. In the beginning, the distribution of costs between departments can be relatively simple by just taking your cloud bill and split it between the different business owners. But with time more decisions need to be made e.g., how do you know which department is using what? how will you distribute hidden costs? and how do you distribute shared costs (shared Azure infrastructure, the FinOps team, and much more) and resources (containers, micro services etc.)?

Getting started you need to work with tags – and we strongly recommend adding virtual tagging to reduce administration. Consider using an external FinOps partner to get going in just a few weeks and avoid the common pitfalls that will make it time consuming to maintain.

Cost architecture – Use the most cost-efficient services that fits your needs

Avoid using unnecessary expensive services/ building blocks. This will save you another 10-15%.

Cost architecture is most likely the last step in the FinOps journey. Surely your team has done some research on this, but cost architecture is an ongoing activity – just like the other FinOps activities.

The reason for needing to do this ongoing is that the cloud has in-build challenges e.g., that the price for one service never drops. Consequently, you need to change server if you want a cheaper compute option.

What you specifically should focus on in your organization varies, but we see over and over again that SQL-servers, VMs, evergreening, IaaS vs. PaaS and Data Analytics are topics where an annual cost architecture review will greatly pay off.

“How do you ensure sound financial governance without creating a bureaucratic overhead sacrificing flexibility and agility? and adding a huge job on your scarce cloud resources? “

If you do these 4 things your bill will be halved compared to if you do not do any FinOps initiatives.

You might have taken some steps already or have not started yet. The most important to do right now is that you get started – 50% off your cloud bill amounts to a lot of money that could be used on other priorities.

The most important piece of advice we will leave you with

Starting from scratch and building your FinOps efforts is complex when FinOps is not your primary assignment. The good news is that your needs will look the same as other organizations, making FinOps an ideal external assignment. That way, you avoid a slow implementation process which only result in paying Microsoft or the other cloud providers more than necessary. With a FinOps partner you get a guide that asks the right questions, and helps you overcome the complexity by learning the simple principles of FinOps.

Our most important piece of advice is therefore to engage with an independent FinOps advisor, that can provide you with both tool and training.

Ready to move on? Have the FinOps journey infographic at hand so you know what to do when, or book a 30-minute inspiration meeting with us.

Get the FinOps journey Infographic

… have the FinOps journey at hand all the time so you know what to do when.

Download The FinOps Journey Infographic

Full overview of the different steps

Full overview of The FinOps Journey

… have the FinOps Journey at hand all the time so you know what to do when.

Three Ways to Save even if Your Azure Monetary Commit is Larger than you Need

The most obvious benefit of entering a Monetary Commit is, that it is often possible to get 3-5% discount on the commitment – meaning you get more Azure bang for the buck.

But what if you have made a commitment larger than your actual need?

For larger customers with a Microsoft EA (Enterprise Agreement) or SCE (Server Cloud Enrollment Agreement), there exists the option of agreeing to a Monetary Commitment for purchasing Azure Services.

#1: Your Azure Monetary Commit may be “True Down Eligible”

It is possible to elect for an annual commitment, that is true down eligible – meaning you can actually reduce it. If your agreement is true down eligible then just move on with that.

If not, make sure that Procurement is aware of this option next time they re-negotiate the EA or SCE.

#2: Buy Reservations for future use

If you have made a single Monetary Commitment, you’re on the hook for the whole amount.

This means that you have to pay Microsoft that amount. However just because you have to pay the full amount, it does not mean that you have to use the full amount during that period!

If you do not use the full commitment, make sure you use the remaining amount to buy reservations to cover future use. What you reserve at the end of year 3 will keep you covered for nearly 3 more years.

This doesn’t improve your cashflow, but it does improve your bottom line, as reservations can be activated and amortized over their full lifetime. This is critical for your CFO to know and understand.

#3: Continue saving

Don’t fall for the sunk cost fallacy, reduce your commitment going forward instead. I know it’s hard, but accept the fact that the commitment wasn’t well aligned with your needs and move on. Fight the tendency to give in to “well I’ve already paid, so I might as well find something to run there”.

There is no reason to just “use it up”, because in our experience, you’ll end up having those services for a lot longer than you planned – meaning you will have been throwing good money after bad, when you didn’t need to. This is most relevant for IT.

This last option will in itself not save you cashflow nor give you any bottom line effect this year – but it will certainly instill the financial prudence in the organization for long term cost efficiency.

New storage Elasticsearch tier in AWS

We all know the story of Goldilocks and the Three bears. One bed was too big, one bed was too small, and one was just right.

Well, until now in AWS for storage, it’s either been too big or too small (or too fast or too slow), with fast(ish) EBS and slow S3 being your choices. This has been a challenge for certain large-volume datasets that need fast performance, at least in the early phase.

An example of this has been log analytics. Something many of our customers use AWS (and specifically Elasticsearch) for.

Now, for log analytics at least, there is a “just right” option with UltraWarm storage tier for Elasticsearch. More about UltraWarm and Elasticsearch here.

Pricing wise, the UltraWarm Elasticsearch instances are about 70-120% more expensive than the non-UltraWarm PAYG instances, and can not be reserved at the moment. However, the UltraWarm managed storage is less than 1/5th the price of EBS SSD.

In other words, you can save up to 39% in some scenarios:

Is this a good business case for you? It depends, but we’ll be sure to include this savings opportunity in your Kostner SAVE analysis, if it does.

Interested to learn more? Book a meeting.

Did you like this article? Subscribe to our blog or follow us on LinkedIn.

A customer in the middle of a cloud journey to Azure asked me this question. They have not yet settled on all architectural decisions for their virtual server environment and they are not running full scale yet. A great question to ask – because one thing you do not want is to limit your flexibility and risk wasting money. After all, isn’t flexibility one of the key things cloud is all about?

Fortunately, the answer is in most cases straight forward:

"Start reserving now!"

There are four main reasons behind this:

The savings on many reservations are so large that payback time is very short. Breakeven is often 5-6 months for a 1-year reservation and 9-12 months for a 3-year reservation.
Most reservations can – at no cost – be changed to other reservations within a similar category of workload.
Reservations for up to USD 50.000 per running 12 months may be canceled.
As Azure fills up, the discount you get for reserving falls over time.

Payback Time on Reservations is Short

Most reservations offer 30-70% discount on an annual reservation.

If we take a 60% discount (typical on many virtual machines) this means that after 5 months on Pay-as-you-Go you will have paid more than the price you will pay for 12 months on a one-year reservation.

Reservations can be changed within the same category

Reservations can be changed within the same category.

This means that if you have made a reservation on ie. compute and you later find out that you will change which servers you are running, then you can change the reservation you made on one type of server to another – at no cost, as long as you are reserving for as much money or more than before.

This is true for most reservations but check this before you buy.
What this really means is that if you expect to use the same or more on ie. “Compute”, then it should not be a problem to buy your reservations now even if all architectural decisions are not made.

Frankly, we almost never experience companies reducing their spending. Most companies start using a new platform like Azure and don’t look back. They will put new things on Azure going forward and migrate relevant workload from on-premises. It is highly unlikely that your cloud spend will be smaller in 6 months than it is now. Even if whatever project you are currently working on should end up not becoming a runaway success, something else will most likely take its place.

Reservations for up to USD 50.000 may be Cancelled

Should the unlikely event happen that you end up with too many reservations this is most likely not a problem. You can cancel up to USD 50.000 per running 12 months. Microsoft reserves the right to charge you a 12% cancellation fee, a fee they are not charging at the time of writing. Why? Because they can resell the capacity at higher unit costs.

Discount levels drop over time

Microsoft has a really good pricing team. They effectively price cloud on a value-based pricing principle. Much like hotels they use advanced yield management algorithms to help drive the maximum revenue possible from the available capacity. What we have observed, is that over time, the discount offered for reserving instances drops. You can see this in FS (old) and FSv2 (current) instances, where the 3-year reservation discount for the old instances have dropped to 50.5%, and the same discount for the new is 64%.

Start reserving now!

Therefore, the answer is so straight forward. And we did not even mention the bonus benefit: you can pay your reservations monthly – it is no longer an upfront payment (although you can chose that, if you have committed spend you need to use, or just too much cash laying around)

The situations where you need to be careful is if you are considering stopping your cloud journey within a very short time frame and you are thinking of making reservations exceeding the 50.000 USD threshold. Again, think across the organization – not project or team-specific.

Actually, we would argue that the decision to start buying reservations is not the tricky one. The challenge will be to find out which reservations to buy – and to ensure that you are covering all areas where you can benefit from reservations. And not buying too many.

Any more questions on how to save on Azure? Subscribe to our blog or follow us on LinkedIn.

Did you like this article? Please share with your friends

You have probably heard about reservations for servers in Azure at this point. You may even be using them to some extent.

What you may not know, is that there are 17 other categories of services you can reserve in Azure right now, promising 30-70% savings compared to pay-as-you-go.

Get our 3 best tips in this blogpost.

In this blog we will share with you the three categories of reservations where our customers could be saving the most money (but are not aware of it):

Linux licences,
SQL Server Elastic Pools, and
SQL Data Warehouse

Linux Licenses

Few customers realize that you can buy reservations for Linux license(we know, we know, not technically a license, but, in practice you have to pay to use a piece of software… so we’ll use that term) to cover SUSE and Red Hat.

An example is SUSE Linux Enterprise Server for SAP Priority. This would be in use for many different SAP use cases, and we see many customers where this saving is applicable.

So how much could you save? A lot. Around 64% actually. 64% of what? Well, for medium server sizes with 5+vCPU the annual savings are about €1200. Per license reservation.

For a typical customer with €50.000 monthly spend, the annual savings are often around €20.000-€30.000.

SQL Server Elastic Pools

A relatively new, and certainly under-utilized, reservation category is SQL Server Elastic Pools. Now not all pools are equal, you are going to need a vCore-type elastic pool. If you are using eDTU based pools today, or are still running stand-alone SQL Server databases, now is a great time to consider making that switch. It is especially relevant if you have a moderately or larger sized SQL Server environment in Azure.

Savings are around 30-35%, depending on region. In real money, even for the smallest pool that is €1500 per year.

Please note that this reservation type is also available for some stand-alone SQL Server databases.

Azure SQL Data Warehouse

A favorite among those who develop Business Intelligence or other Analytics solutions in Azure, is the use of Azure SQL Data Warehouse. I use that name here, as that is the name you will find in the Azure Portal, in marketing material and on Microsoft’s website, it is now known as Azure Synapse Analytics.

What has yet to become a favorite, but certainly should be, is using reservations to cover your Azure SQL Data Warehouse consumption.

Reservations are bought in blocks of DW100c’s – an hourly consumption metric. So, if your services use 1500 DW100c’s continuously, you would reserve 15 units of DW100c. A pitfall here is that many customers vary the size of their instances over time – and spin up additional instances in some periods. You will want to control for this when making reservations.

Learn more about Savings in Azure

Next really depends on the purpose of your efforts:

If you want to become an expert and learn all the details of how to save on Azure reservations Microsoft provides you with a thorough description.

Or learn more ways to save on Azure by joining our free webinar.

Join Free webinar

… and lastly, if what you are looking for is a way to save on Azure, with limited time spent from your side then book a meeting and learn about how we might help you.

Book a meeting

Remember

If you want to receive updates about how to save money on IT, then get updates. Subscribe to our blog or follow us on LinkedIn

Did you like this article? Please share with your friends