Seven ways to save money on Microsoft Azure

To get right to the point, we’ve helped customers save as much as 25% annually on their Azure costs through our resource optimisation and license optimisation review. It’s one of the most valuable services we offer to our customers because of the significant savings we usually uncover.

We work with Azure all the time. We move customers from on-site infrastructures to the cloud, help set up hybrid clouds, or shift them from one cloud to another. So we know just how tricky it is to get things right, and how much time, effort and knowledge is required to maximise the value of your Azure platform.

You’d imagine that you simply need to purchase the same virtual machine capabilities in Azure as the physical machines you are already using. And that seems logical, right?

However, the ‘carbon copy’ approach isn’t necessarily the right one. For example, if you had a physical device running eight cores, with 32gig memory and a terabyte of hard drive space, and you replicated it in Azure, you’d end up buying and paying for more capability and power than you need.

So, what do we do that you (likely) can’t? When we perform a resource and optimisation review, we plug our assessment tools into your tenancy and run them for a week. The resulting report may uncover that of your ten virtual machines, devices 1, 2 and 3 are running at only 20% capacity through the entire week. So you are paying for 80% of computing power, which you don’t need.

Tip: We can go through the same exercise to look at your CPU, memory, and disk utilisation, so we can buy smarter on your behalf.

Another trap is not understanding the performance levels of the Azure compute layers. We frequently see customers purchasing premium storage where they could have more appropriately chosen, and paid considerably less, for standard storage.

There are several configuration options in Azure which impact the data replication points (including LRS which replicates your data at the same data centre, through to GRS which replicates your data in different geographic locations). It’s confusing, so we’re never surprised when customers select the wrong storage. You have five types of storage to choose from, each of which then offers a further five options for redundancy types. It’s easy to end up paying for additional storage and points sitting in Azure which you don’t need.

And if you make the wrong decision from the outset? For example, choose the most expensive storage option (ZAGRS), and realise your mistake six months down the track? You’re stuck with it until the end of your license period.

Additionally, Azure offers two types of disk storage: Standard hard drive and SSD hard drive (flash memory). Choosing the wrong one can cost you dearly. For example, adding 512gb disk to a standard hard drive, it will cost you $34 a month. But if you add the same 512gb to an SDD drive, you’ll pay $52.

As Azure evolves, Microsoft brings in more computer types - and different pricing to match. If we compare the F8 server (eight cores, 60gig memory and 128gig of temporary storage) with a D8, which has double the memory and the same CPUs - it would cost you $20 less a month. Or if our review shows you are underutilising your device capacity; we can double your savings again with an F4 with its four cores and 60gig of memory.

And a trick for beginners: If you’re committed to Azure for the long-term, you can reserve the storage you need and earn a discount from Microsoft.

During the excitement of moving to your new virtual machines, it’s also easy to forget that you can use your existing on-premises port licenses and the software assurance component of your agreements to reduce costs. You already own the software and may use it on-premises or on Azure, so don’t get caught out by paying for it twice! The same applies when you purchase a compute layer in Azure. Make sure you understand what you are buying and don’t replicate what you’re already entitled to.

Tip: If you’re planning to move to Azure, check with your partner that your software assurance licenses are up to date so you can transfer them over.

Azure also offers various tiers of performance of virtual machines, which can get tricky. You may need multiple environments for pre-production and production. In Azure, you can choose which level of computing power will meet your requirements. For example, for a pre-production environment, you don’t need high-speed performance, because you’re only setting out to prove that the solution will work - not how quickly.

As with storage, we can optimise your compute environment so it’s not overcooked. You can achieve some substantial discounts (from 32% up to a whopping 57% discount) if you make a one or three-year commitment.

As an example, if you selected an F8 virtual computer from the Australian East data centre, it would cost you $658.97 a month. A one-year contract would bring the cost down to $511 a month, or $432 monthly with a three-year contract. You’d respectively save $1775.64 and $2723.64 a year.

Tip: If we select and build out a standard rather than a high-performance virtual machine for your pre-production environment, you’ll save even more.

The snapshot feature is a nice-to have but easy-to-get-caught-out-on feature. At $132 a month per server, the snapshot feature enables you to rollback your environment to its previous state if you have an unsuccessful upgrade. Which is very handy.

But it’s easy to forget to turn it off. And when you are running it across several virtual machines, it soon adds up.

As mentioned earlier, we often see customers replicating their on-premises environment on Azure. However, with some knowhow, you can take a more modern approach. Using app services, you can pay to use the application on the machine, instead of for the entire server. In effect, it’s an IaaS model, more commonly called serverless technology - where you only pay for the service, not the server!

App services also allow you to configure services to turn off automatically when inactive (for example, overnight). When you buy a virtual machine, it runs all the time, and you pay for every minute of your contract. In comparison, you only pay for serverless technology as consumed. So by automating the service turnoff, you can reduce your costs to the minimum virtual cores that you have selected in your solution and only pay for the active times.

Tip: Server-less technology offers great bang for buck, as you can use better applications. As you’re only paying for when you use it, it becomes better value than an always-on, but less functional application. And if you have spikes in usage, you can scale up your serverless delivery for far less than permanently adding the virtual servers you’d need year-round.

Now, this may sound strange, as you would imagine that all things Azure are created equal. But costs often vary between data centres.

All data centres are built differently, and the costs to run them vary according to local wages, construction and maintenance, and utilities like water and power, and fuel for backup generators. Microsoft strives to achieve a low to zero carbon footprint in all their data centres, so they are careful to choose the right cooling mechanisms for each. However, depending on the location of data centre, the cost to control temperature can differ. You can imagine the expense of keeping a data centre cool in Dubai vs Ireland.

When you deploy your services in the wrong data centre, you can suffer from degraded performance due to latency. And that cuts both ways.
Hosting your data offshore can also present governance issues, and require you to migrate your Azure environment from one country to another.

Tip: Do due diligence on which data centre you choose. It's costly to repatriate your data!

Investing in a resource and licensing review will pay dividends. We have a team of experienced Fusion5 consultants who specialise in helping you maximise the value and performance of your Azure infrastructure. They come ready-prepped with years of knowledge, a deep understanding of Microsoft licensing and configuration, and a desire to help you save.