Views, News & more
Organisations that switch to the big public clouds often find themselves paying more than expected.
If you're in this situation, here are some ways to cut your public cloud bill without materially harming performance:
Check actual RAM usage and CPU usage at a variety of times (during the working day, during backups, when generating reports etc.) Downsize to smaller instances where feasible. If that causes issues, just upsize. The cloud’s flexibility means you don’t have to get sizing right first time.
Pause them for a week to see if anyone notices. If not, save the VM image to disk, disassociate any attached storage and IP addresses, then delete the VM.
If a VM is rarely needed, consider scheduling or scripting it to be spun up when needed and spun down when not. You may also need to spin up such VMs occasionally for patching.
If your VMs require lots of RAM or CPU cycles relative to other resources, check if less popular instance types might be a better fit in terms of cost and performance.
Review the data your organisation is storing. Do you really need to keep stuff from ex-employees who left long ago and folders that haven’t seen a new file in over six years? If not, delete them. Unnecessary storage doesn’t just cost money directly; it also increases your backup costs.
If you need to keep old data, check you’re storing it on an appropriate storage tier. Cloud services often have cheaper storage tiers for data accessed infrequently, especially if you can tolerate data retrieval delays. Second-tier cloud providers often offer materially cheaper object storage.
Once you understand the minimum level of cloud resources you need and how that’s likely to grow or shrink, you can decide whether to commit/pay in advance in return for substantial discounts. Spot instances offer a cheap option for non-urgent workloads that can be paused at short notice.
This has serious downsides for security and stability. However, if you have lots of VMs that don’t require many resources but need to burst to serious usage at no notice, consolidating to fewer, bigger VMs can cut costs. There are better alternatives, based on VMware. For more information, read our full guide to cutting cloud costs.
Start with 'just big enough.' You can increase the size of your databases later.
It’s not just slowing down user responses; it’s forcing you to host on larger VMs than might otherwise be the case. So, fix the worse of your code. Consider whether cached data could be substituted for live data in some cases. Leave large-scale sorting & filtering to database systems.
If your website gets lots of traffic and largely consists of infrequently updated non-personalised content, consider using a content delivery network (CDN) such as Amazon CloudFront to speed up responses and reduce & flatten the demand on your web server, allowing for it to be downsized.
If you’re already paying for Microsoft 365 for access to Word, Excel, PowerPoint etc., subscribe to a Microsoft 365 tier that comes with hosted Exchange and SharePoint. That way, you can decommission your old Exchange server and use SharePoint sites as your shared network drives.
This won’t cut your total bill but may cut your cloud hosting bill. SaaS/PaaS services can simplify IT, leaving others to patch and update those apps and take care of the underlying platform. Resilience & scalability may improve. That's great if your on-prem systems are reaching end-of-support.
If you’re out of contract, seek like-for-like quotes from multiple cloud providers. Make sure you’re clear about what’s required in terms of support, backups and monitoring.
This can cut cloud storage costs significantly as most data compresses significantly.
Already have a licence for the software you’re planning to run? ‘Bring Your Own Licence’ could save you paying twice. Do you really need Windows Server, Internet Information Server or SQL Server? Or could Linux, Apache/NGINX and MySQL suffice? Consider switching to open-source.
It’s easy for hyperscale cloud costs to drift upwards. Set up alerts to help you spot when resources are ‘temporarily’ provisioned and mistakenly left running.
If you spend a lot on your cloud, consider allocating your cloud costs to different departments, projects and customers. It’s a hassle, but it will help you discourage excessive consumption. You may be able to ‘tag’ resources to keep track of which department or project should get the bill.
For further tips, check out our hSo's free guide to cutting cloud costs.