Deciphering the Azure Saving plan

Current scenario

Until now in Azure, if we talked about how to save costs in computing services, we had two alternatives (without going into software discounts):

• Azure Reservations: They help us save money by booking a particular VM size for 1 or 3 years. The cost savings can be up to 72% (official figures from the manufacturer, which in my experience of 40% has not passed) compared to Azure prices in PAYG format. In this case, when we made a reservation of an instance, it does not affect the state of our resources, but we made the reservation against a specific size, obtaining the discount automatically if it matches our resources.
• Spot Virtual Machines: This type allows us to have a machine for computing at a lower cost than normal, but with one condition: We do not have SLA, when Azure needs computing capacity, the first thing that will be rescinded are the Spot type machines, with which we would be left without the ability to have computing resources for this type of sizes.

But during Ignite ’22, a third avenue for computing resources was announced: Azure Saving Plans

And what does it provide me?

It allows us to save computing costs based on a fixed price per hour. In this case, a Saving Plan can save up to 65% the price compared to an Azure price in PAYG format (manufacturer figures, which in my experience of 30% has not exceeded), always depending on the term we choose (from 1 to 3 years).

And what is the main difference?

Basically in the way of reserving the resource, if for an Instance Reservation, for example, we reserve a size D4v4 in West Europe for one year, with Azure Saving Plan, what we do is set a fixed spending rate for a certain term (from 1 to 3 years without the possibility of cancellation), so that any computing resource that falls within the scope we have chosen can make use of that commitment and This saves us computing money from these resources.

How does it work?

Basically, we have to specify the amount of fixed money we want to spend per hour of computing, and automatically, all the resources that are contained within the scope of creation

Therefore, it is extremely important to keep in mind that this type of solutions do not fit with everyone, since not all of us have a large amount of computing resources that involve a fixed cost for our organization.

Likewise, we must specify how long we want to have this commitment (1 or 3 years) and the form of payment (monthly or annual)

When trying to create an ASP, the portal will offer us different alternatives to configure our ASP depending on the computation consumption we have, from the most conservative to the most aggressive strategy (although manually, we can also configure how much we are willing to pay per hour)

Once we have created the ASP, the party begins: How do I know that I am applying the ASP to my resources? The answer is simple, you must trust 😛

A sample of how it works is the following image:

If you look, the green line represents the amount of money I am paying in a fixed way every hour (remember that this is 24×7, so if we put € 5 / h they end up being approximately € 3600 per month), whether or not I use computing resources.

This last sentence is very important, whether or not you use resources, what does this mean? That, if I use 100% of my computing resources, and the price / hour is less than those € 5, I will pay € 5 / h yes or yes. On the other hand, if my 100% of computing resources / hour is greater than those € 5, I will pay € 5 in fixed format (which already contains a certain discount), and the remaining € 1, I will pay it at PAYG price (remember depending on the contract I have).

So here, we enter different price scales:

1. Scenario 1: In a certain time slot, I go below my set price à I pay my price per hour
2. Scenario 2: In a certain time slot, my computing consumption is what I have set in the ASP à I pay my price per hour
3. Scenario 3: In a certain time slot, my computing consumption is greater than the ASP created à I pay my price per hour + PAYG price not covered by the ASP

This is important to understand, because savings are automatically applied every hour, regardless of region, instance series, or OS.

What resources are contained in this type of solution?

As I write the article, different Azure resources are coming into play such as:

1. Azure VMs (excludes A, G, and GS series)
2. Container instances
3. Azure Functions con Plan Premium
4. Azure App Service with Premium v3 or Isolated v2 Plan
5. Azure Dedicated Hosts

This does not mean that other resources will be included in the future, but I do not have more information.

And can I combine it with instance reservations?

Yes, without problem, in fact it is the most suitable formula to save costs, in this case, the instance reservations would always enter first, and everything that does not cover the instance reservation, would be subject to be covered by an ASP:

As we can see, everything that is not covered by the instance reservation or an ASP, would be paid at the normal compute price that we have established in Azure (here it will depend on the type of contract we have with Microsoft EA / CSP / PAYG)

And this ASP thing appears in Azure Advisor?

Yes, it should already appear in the Advisor as a saving measure for the compute services contained in Azure along with the reserves of instances

We must even realize that this option already appears in the Azure calculator:

Once I have made a commitment with ASP, do I have the possibility to cancel and/or change it?

No, it is not possible to cancel an ASP commitment, or change it for another, we will have to endure 1-3 years what we have configured, and if we fall short, we will have to configure a new ASP to cover the new demand (with the increase in time of this new ASP that supposes)

What you can do is switch from an Azure Instance Reservation to an Azure Service Plan Self-service trade-in for Azure savings plans – Microsoft Cost Management | Microsoft Learn But not from ASP to RI.

Any recommendations for creating an ASP?

My personal recommendation is to always go to a more conservative configuration, more than anything because of the non-possibility of being able to cancel this type of commitments, so this will give us the opportunity to “play” with other configurations.

Summarizing

Reservations only apply to computing resources that have been identified and to a specific region

Azure Saving plan applies to all compute resources that are contained within that scope, so they provide us with greater flexibility and automatic optimization against reservations.

When to choose one or the other?

1. For compute resources with dynamic loads: Azure Saving Plans
2. For resources that are stable over time and run continuously, or don’t think about resizing: Azure Reservations

There is no one-size-fits-all formula, but the FinOps perspective is like this 😊.

Additional information about Azure Service Plan at: What is Azure savings plans for compute? – Microsoft Cost Management | Microsoft Learn

Advertisement

Best Practices about how to cut costs in Azure

Introduction

There is no one sizes fits when it comes to Azure and cost optimization, but the focus of this session is to explain some tips & tricks during my daily life as a Cloud Solutions Architect​

Some general tasks can be done monthly/quarterly to be sure that you Azure environment is up to date, taking into consideration that the optimization and your business run are the most important things here​

Be advised that not all the things that can be done in Azure are being covered in this post, probably because at the time of the writing I didn’t have to

Why this post?

Every design in Azure has cost implications, before architecting something, we must consider the budget that we will need for the Project itself, taking into consideration thinks like:

• Identity different boundaries for scale up
• Redundancy
• BCP taking into consideration the cost of the solution
• Design and set up scalable architectures, focusing on metrics & performance
• Start small and scale out as soon as the required performance needs it​ (I really love that one)
• Choose PaaS and SaaS over IaaS, pay only for what you use as a consumer​
• Always, monitor, Audit & optimize the cost related

Ok I get it, but what are we going to cover?

For the next minutes I will explain some guidelines about cost optimization, in particular for the following topics:

• Use of ARI
• Use of Dev subscriptions
• Optimal use of Azure App Services​
• Optimal use of Auto-Scale in App Services​
• Azure Data Factory Failed Pipelines
• PaaS SQL Optimization
• Cosmos DB
• VM Right Sizing
• Azure Hybrid Benefit​
• Blob Storage Lifecycle​
• Networking
• Clean Orphan Resources​
• RIs​
• Use of Log Analytics
• Use of Azure Advisor
• Cost Management Preview (ACO Insights)
• Azure Governance Dashboard
• Closing

Before starting…

Before starting this post, I would recommend to create an Azure Inventory from your environment, with this tool, it is pretty simple: https://github.com/microsoft/ARI

And as you can observe, it will give you a great overview of what type of resources are you having, which use, locations, etc…Also, some of the sheets can be used to optimize your Azure Cost environment

Also, another tip that I want to give, is you can start your journey to Cost Optimization with a self-done assessment, but it can give you some guidelines about where are you: https://docs.microsoft.com/en-us/assessments/

Use of Dev Subscriptions

Using the top-to-bottom approach, the first thing to pay attention to is Azure Dev/Test subscriptions, which are applicable for both enterprise and pay-as-you-gooffers. By placing your dev resources in those subscriptions, you will get lower prices for most common Azure services for the cost of excluding them from the regular vendor SLA commitments. 

Optimal use of Azure App Services
First, check that standard Plans and Premium plans has an associated application​

I have seen a lot of empty App Services Plan, which leads to unnecessary cost to the customer, remember that having a right governance in your subscriptions it is algo a cost measure.

Another thing that I tend to do is to check the metrics for the plan, and check if are being used properly (scale down in case is needed, but remember the features needed in each case)

Optimal use of Auto-Scale in App Services​

In my case, I can be able to scale down my resources, but first check features between standard and premium plans (or even between std o premium!). Also what it can be done is to scale up/down based on a schedule https://techcommunity.microsoft.com/t5/apps-on-azure-blog/azure-app-service-automatic-scaling/ba-p/2983300

Very useful for those workloads where we know that only are needed in certain periods of time

Azure Data Factory​

Review the failing pipelines​, if a pipeline is constantly failing to run, probably it will impact into the cost of your resource, so take action

Again, I have reviewed a lot of Pipelines in DF which are continuosly failing… take care of that as well

Paas SQL Optimization​

With monitor, check if the database needs all the DTUs provisioned, one thing I love to do, is to play with the different available plans for the SQL, if you’re running a version with a lot of DTU’s, implement a runbook in order to reduce the plan when you don’t need it​, for example, you can use: GitHub – francesco-sodano/azure-sql-db-autoscaling: This ARM Template deploys an Azure SQL Database with DTU Consumption plan (with a new Azure SQL Server) including all the resources required to perform Auto Scaling (scale up and scale down) based on Metric Alerts using a function app. Again, very useful for those workloads where we know that only are needed in certain periods of time

Focus on those DBs with 40%-80% of the DTU capacity​, those are the most imporant to be scale up

Check if you really need the Georeplication, probably you don’t need to replicate your DB across regions (important point!!!), remember the first bullets of this post, we need to start small and then plan big, if we start to put all the georeplication modes to DBs that are not being use or for those in test, you’re wasting your money

Cosmos DB​

With the help of metrics, review the use for the correct size & throughput ​

Consider autoscaling for those type of resources (it avoids consuming unnecessary resources)​

Consider serverless options for Dev & Test environments or those environments where intermittent traffic is it used: Consumption-based serverless offer in Azure Cosmos DB | Microsoft Learn

VM Right Sizing​
One thing I love to do, is to shutdown VMs based in a Schedule​

The Schedule is set up with a tag in the resource, and the operation is done by an Automation Account (it could be a Logic App as well)​. For example, I love to use the following script:

Scheduled Virtual Machine Shutdown/Startup – Microsoft Azure | Automys

You can setup the following tag in the VMs

And your VMs will automatically shutdown and start in the configured schedule, which for those test and PRE environments where Azure Reservations does not fit, are simply great, you will save a bunch of computation hours with this simply script

In order to cut costs, we can use spot VMs for non priority tasks (it helps to save some money vs other azure VM sizes), you can get more info in: Use Azure Spot Virtual Machines – Azure Virtual Machines | Microsoft Learn

Get rid of those old VM sizes

One thing that I do for all of my clients in order to optimize cost, is to check which version size are running for the VMs, this can be extracted from the ARI (remember the first tool):

Why? Because as you probably know, Microsoft is always optimising hardware in the Datacenter, so they are pulling new version of the VM size, so what’s the point? the older is the VM size, higher VM cost, so check out if there is any new VM size, and you will be able to save some money from each VM size.

Imagine that you have 100 VMs running in a v2 series, and changing from v2 to v5, represents a change in cost of 20€/VM/month, so in total the save is 2000€/month with only changing the VM to a newer version, not bad uh?

Azure Hybrid benefit

First question is: Do you have a software assurance with Microsoft? If the answer is yes, don’t waste more time and money, and apply it to your Azure Resources, it Will help to sabe up to 40% in cost (for VMs and SQL)

If you want to know how much you can save with this, you can use the Azure Calculator for this purpose: https://azure.microsoft.com/en-us/pricing/hybrid-benefit/#calculator

Storage Lifecycle

• Everyone loves to sabe data, but what can be done to save some money? Use lifecycle J Optimize costs by automatically managing the data lifecycle – Azure Storage | Microsoft Learn
• Transition information to a cooler Storage tier, to optimize performance and cost
• Delete information at the end of the lifecycle

With this procedure I was able to save a lot of money in a recent IoT Project, all the information was stored in blobs, but once a certain period of time passed, we moved the information from one tier to another in order to cut storage costs

Networking

Check out costs related with networking, it may scary you​

You Will need to identify which applications are using most of the egress bandwidth and review & redesign your infrastructure accordingly​

Check which gateways are not being used, probably those which have a throughput lower tan 900MB/day​

Check you Azure Express Route Circuits, probably the first provision of the circuit was greater than needed

So, check Azure Monitor: Monitor – Microsoft Azure

Clean Orphan Resources

Are you sure that everything that you have in your subscription are being used? Use this workbook and take action in your subscription: Azure Orphan Resources (microsoft.com)

Save Azure costs deleting those unused disks, Public IP’s which are consuming Storage and account cost (remember that in Azure Advisor we have these recommendations as well):

I’m sure that you will save a bunch of €€€ with this procedure

Use of Log Analytics

If you’re using Log Analytics to monitor your Azure Resoruces, you should add a Daily Cap into your Log Analytics Workspace: https://learn.microsoft.com/en-us/azure/azure-monitor/logs/daily-cap#view-the-effect-of-the-daily-cap

Also a few tips:

• Use Azure Monitor Agent and Data Collection rules over Log Analytics agent
• Set retention per table and leave the workspace retention to its default
• Set archival tier per table – To meet certain compliance rules, you may need some of the data available for a longer period of time
• Configure diagnostic settings with only the logs that are needed and used

Use of Azure Advisor
I must admit that I’m a fan of Azure Advisor, for any Project that i have, always i tend to revise Advisor​ in order to cut Azure costs

It helps to detect if a Virtual Machine runs on a VM size GREATER than what it needs (based on CPU utilization under 5% in the last 14 days). If the Azure Advisor reports an overprovisioned machine, you need to investigate its use and resize it to a more suitable size.​

For this VM rightsizing purpose, I also use a script from Jos Lieben, which helps to put your underused VM in the right size in terms of load: Automatic modular rightsizing of Azure VM’s with special focus on Azure Virtual Desktop | Liebensraum

Reserved Instances

Reserved instances allow us to reduce cost, there are a lot of resources that can be reserved, take them into account when you’re designing your infrastructure

As you can see, there are a lot of Azure resources available to be reserved, make use of them 🙂

In Azure Advisor always recommend to reserve instances of our resources, don’t forget it​

Azure Budgets

Send notification when a certain amount of money is spent, this can be set at resource group or subscription level, and for example email to the application/subscription owner

Azure Cost Management

Remember to keep a closer look to the latest updates from Cost Management: https://azure.microsoft.com/en-us/blog/microsoft-cost-management-updates-november-2022/ I’m sure that you’ll take profit of those new features 😉

Insights, is the new feature of ACM allows us to have some insights about our daily spending in Azure resources, we can detect what is a tendency, and what is a cost anomaly in our subscriptions

Azure Governance Dashboard

If you want to deploy a High-Level Visualization in PowerBI of your azure resources, you can implement the CCO Dashboard from GitHub: https://github.com/Azure/ccodashboard

I know that this is more related with governance, but it helps to have a bird’s eye into the different resources and Azure subscriptions.

PRO TIP

If you really like those cost recommendations, there is a toll in github: https://github.com/helderpinto/AzureOptimizationEngine which can enchange the Azure Advisor recommendations and help you to optimize your environment

Closing

That’s all, probably some of the recommendations are already being followed by you, but I hope this post was interesting to you 😊

Till next time, merry Christmas and happy holidays!

FIVE KEY QUESTIONS TO CONTROL CLOUD COSTS

Today I’m not talking regarding Azure, I’m talking generally in the Cloud, so… the case here is: you’re ready to modernize your infrastructure to truly benefit from the cloud and avoid sprawl, poor performance, complexity, and sky-high subscriptions. Here are the five key areas you need to plan things out.

1. What problem are you solving? The business case behind your infrastructure must be understood, with input from key stakeholders. How will the cloud help you deliver that value to your users or customers? How will you benchmark and improve on that goal? Common goals for cloud infrastructure are faster service delivery, lower operating costs, agile deployment, simplified user experience, modernization, and resilience.
2. What base infrastructure components will be used across your stack? Your subscription model, identity management, network, storage, backup, and other key components should work with any given app or VM that might be attached to your subscription.
3. Where can you automate? Automation and standard procedures are key to realizing the full benefit of the cloud. This helps minimize costs by automatically provisioning and decommissioning and implementing configurations without intervention. Infrastructure as Code will only continue to be more vital to cloud management.
4. Who is responsible for each piece of the stack? As you embrace more and more components delivered entirely as a Service, you must understand which entity is responsible for their management and potential failure. Obviously the underlying hardware is not yours to command, and a hardware-level failure would fall upon the service provider. But issue-resolution can be a difficult gray area. Be prepared to work with the cloud provider and have processes in place for issue tracking and resolution as well as version control and backup/restore.
5. How are you managing cloud governance? Depending on your management models, it may be much easier for users within or outside the IT department to provision their own resources. It was much harder to order, install, and configure a physical server than to simply click a few times to have a virtual one ready to go. You must be ready to protect and secure your data from inside and outside risks, while also reining in sprawl and maintaining compliance. Automation tools and code are a good place to start with governance and compliance.