Cloud computing has been one of the most significant technological revolutions of the 21st century. Enterprises are now having to completely transform the way they think about budgeting for the provision of additional infrastructure – you’ll also need to adapt to keep pace with competitors.
In their haste to adopt cloud technology, many IT buyers aren’t fully developing their understanding of how the cloud’s flexibility, on-demand availability and infinite scale affect the provisioning process. This has led to many capacity planning mistakes, resulting in spiralling AWS cloud costs and inefficient performance.
In the last year alone, businesses are believed to have spent over $10bn on wasted cloud infrastructure. By 2020, it's estimated that cloud spending will grow to around 25% of the average business’s IT budget, so the rampant wastage of cloud resources only looks set to increase.
Clearly, legacy provisioning processes don’t fit the new cloud computing paradigm, but how are enterprises changing to meet the challenges of controlling cloud costs and reducing this significant waste of resources?
Only Provision Capacity That's Needed
To prevent hardware limitations from getting in the way of business growth, it was previously common for IT buyers to purchase additional infrastructure on top of their current business requirements – effectively buying space to meet future needs.
This made a lot of sense when dealing with physical servers because provisioning extra infrastructure was a lengthy and expensive process. Hardware needed to be purchased outright, which meant large amounts of capital expenditure (CAPEX) and waiting for financial approval from the board could take upwards of a year.
Unfortunately, not all IT buyers realise this approach isn’t necessary now that small amounts of additional cloud capacity can be turned on at the touch of a button and the cost incorporated into their monthly operating expenditure (OPEX).
Better cloud cost management has become of paramount importance for enterprises to ensure that they are able to prepare for peaks in cloud usage and business growth without paying for wasted infrastructure.
In more complex systems you might find that spare capacity is caused by underlying performance issues. Often, this is where large savings can be made as cloud infrastructure is significantly downsized. To achieve these savings you'll need to carry out extensive performance testing to identify areas of your systems that are fit for optimisation.
Unlike some of the methods we outline in this post, tackling performance issues in complicated systems is a very complex project that can't be done without the use of intelligent built-for-purpose tools.
Improve Code Efficiency Across Your Systems
Most inefficient code issues in your systems won't individually have a noticeable effect on your performance. However, with a recent study revealing that the average system contains nearly 2,000,000 code instances, if just 5% of your code isn't optimised the cumulative effects will certainly make themselves felt.
Code inefficiencies can result in a number of undesirable behaviors including:
- Transferring large amounts of data between a database and an application
- Making a large number of data transfers between a database and an application
- Searching through vast amounts of data
- Putting too heavy a load on the database
Ultimately, these will all lead to your business using far more capacity than would be required if all your code was optimised for efficiency. Again, extensive performance testing can identify any shortfalls in performance and inefficient code issues.
Use Auto-Scaling Correctly
For many businesses auto-scaling is very simple. When they need more cloud capacity to meet increased usage demands it's automatically provisioned and they pay for the extra infrastructure they use.
It's clear how this would have its advantages, but auto-scaling without the proper cloud cost management can be wasteful, inefficient and costly. It's important that enterprises better manage their cloud auto-scaling or they risk:
- Users experiencing reduced performance, for example slow speeds and being unable to access your website or app, which will undoubtedly have a negative effect on your reputation.
- Having an abundance of capacity after the peak in usage has passed, leaving your systems using that extra power to perform everyday tasks inefficiently, like a car going down a motorway in first-gear.
- Triggering higher prices for additional capacity or an obligation to provision much more than you actually need.
Instead of relying on auto-scaling as a contingency plan when experiencing significantly increased usage, enterprises are slowly realising they can use auto-scaling to control cloud costs and ensure high levels of performance.
Better Housekeeping
Embracing cloud cost management means achieving a level of insight into cost spikes that simply wouldn’t be possible if relying on auto-scaling. For example, you’ll be able to tell the difference between instances of increased traffic from inefficient micro-services or superfluous storage.
This empowers you to contact individual teams to let them know how they can help control cloud costs by being much more mindful of where they’re overusing storage or under optimising system processes.
Reduce Idle Capacity
One of the simplest ways that cloud cost management is helping enterprises to reduce wastage is encouraging employees to engage in better cloud capacity management – the simplest form of which is just turning off infrastructure when it isn't in use.
Obviously, this doesn’t mean turning off websites and apps at the end of the day, but if a development team uses a virtual server for building and testing software, it’s not going to be needed over the weekend. Likewise, if there are server environments that are no longer in use, you should ensure they are safely removed.
Switch to Reserved Instances
For many businesses cloud cost management means treating cloud infrastructure as an on-demand resource and only using what they need. However, for some enterprises (particularly those whose usage isn’t subject to significant peaks or troughs) the opposite is true.
Reserved instances are when a business pays in advance to ensure a particular amount of cloud capacity will always be available to them for the duration of the contract, regardless of how stretched the provider’s resources may be.
Normally, you’ll pay a lower cost when purchasing reserved instances (as much as 75% less in some cases) and many providers offer discounts if you pay upfront for the duration of your contract. Of course, businesses pursuing this strategy will need to ensure they’re saving enough money to cover the costs of paying up-front for infrastructure they might not need if usage drops.
Working in-house to better manage your cloud costs help you address some of the "low-hanging fruit" optimisation activities. However, extensive cloud cost management requires a substantial and sustained effort by an enterprise, with the majority of savings achieved courtesy of in-depth analysis of infrastructure usage, regular optimisation activities and close monitoring of performance.
Obviously, this isn't always possible in-house and some of the measures outlined above can have negative effects if not properly planned and executed.
By working with an experienced cloud partner you’ll gain a much deeper understanding of your cloud spend and how to reach the most cost-optimal deployment that meets the specific needs of your organisation – ensuring that every decision is based on hard data.
They’ll also be able to help you identify situations in which you do need a bit more oomph in your infrastructure, for example, if you need to get a new product out on time or quickly fix a bug, it can be best to optimise for speed, rather than affordability.
To learn more about optimising your AWS costs to control you cloud spend and reduce waste, download our new control cloud costs whitepaper.
