Without close monitoring of expanded use, your business could find its cloud bills growing each month. David Penny, Director of Cloud Services at Opal Wave, takes a look at some specific and practical tips for saving money on cloud costs when considering bandwidth, compute and storage, cost optimisation processes and more – all of which can allow you to have a more cost-effective solution for your entire IT estate.

Many organisations have moved their workloads to the cloud as this was supposed to save money, however, users may now find their cloud bills growing and it’s not always clear why this is or how they can reduce their future costs without affecting their business operations.

Reduce Cloud Costs

1. Bandwidth Costs

All public cloud providers charge for data egress, which means you don’t pay for sending data to the cloud, but you do pay for extracting any data from the cloud. These costs can be quite significant, as NASA found out in their recent audit when they underestimated their $30m per year bandwidth costs.

It might also be useful to understand how public cloud providers charge for bandwidth. In the telecoms world, connectivity and bandwidth is provided based on data rates, eg. 100MBps. There may be contention rates if this bandwidth is shared with other parties or no contention if these are dedicated connections. However, in the public cloud world, data egress is charged on data volume, not data speed. There is a free allocation that permits a small amount of data egress, for example with Azure the first 5GBytes transferred per month is free, then you pay excess egress charges for data over this amount.

The problem occurs with public cloud when servers send or stream a lot of data to users continuously – these charges can become quite significant. As a comparison, a dedicated 100Mps telecoms line running 24 hours a day could transfer 32TB of data over a one-month period. The same volume of data as a data egress charge from Microsoft Azure in the UK, at their current list prices, would cost over £2,000 per month. To make matters worse, from the 1st of February 2021 Microsoft will also be charging for Availability Zone Data Transfer bandwidth, which means users are charged for the actual data transfers that occur to replicate data from one Availability Zone to another, rather than just being charged a flat fee for using the service, as it is today.

Businesses should consider alternative connections or locations for these heavy data transfer servers; dedicated connections such as Azure Expressway which can provide a more cost-effective solution for larger organisations, or by utilising third party network providers such as Megaport, that charge a smaller fee for using their network that is already in place. Organisations should also consider using private cloud as well as public cloud particularly where servers have very high data connectivity requirements that can be located in private data centres co-locations where they can benefit from more cost-effective telecom links that offer unlimited usage to stream data.

2. Compute and Storage

Each cloud vendor offers a range of different instance types and storage services aimed at a variety of use cases, cost requirements and performance expectations. When virtual machines are built, designers often select premium storage to aid performance. In many cases, lower tier of data may be sufficient for certain applications, which can be tried and tested using stress testing methodologies.

You can also use life cycle management policies to help lower storage costs. For example, it might make sense to move your older data to lower-cost archiving and cold storage services within the same cloud platform or it may be more economical to transfer it to another cloud or location altogether.

3. Reserved Instance Discounts

Organisations often move to cloud because of the flexibility and scalability and initially choose instances that don’t have any commitment, such as Azure “On-demand”. When cloud is used for development this flexibility is perfect. VMs can be spun up and used as and when required, only to be turned off afterwards and the cloud provider will be only charging them for the hours that were actually used.

However, for production workloads that are on 24 hours a day, it doesn’t make sense to pay this ‘flexibility premium’ when used for long continuous periods. All public cloud providers offer significant discounts on their instances when longer time periods are committed. The amount of discount varies between data centres and on the age of the solution, but these discounts can be huge.

Azure UK, for example, on their older D series of VMs offers a 55% discount for a 1-year commitment and 70% discount for a 3-year period from the normal on-demand prices. Later this year Microsoft will be launching an even bigger discount rate for even longer 5-year reserved instances on many VM Series.

4. Cost Optimisation Process

It is recommended to look at the specific cost optimisation options available from your cloud provider: Microsoft Azure, for example, has its own ‘Cost Optimisation Process’ which can provide recommendations on right sizing or even shutting down VMs that are not efficiently utilised. You can also use Azure Monitor to send alerts to the administrator about various factors, for example if a CPU utilisation is below a certain percentage. Following these alerts, you can then decide to take action manually or even run Azure automation routines to take appropriate actions – Microsoft Azure Recommendations reports can even provide estimated potential annual savings.

5. Multi-Cloud Strategy

In the future, businesses should adopt a multi-cloud strategy to avoid being tied to any one vendor and get the overall most cost-effective solution for their entire IT estate. The best type of cloud will vary across an organisation’s cloud environments dependent on which parameters are considered to be of most importance for that particular app or workload. A variety of factors might be considered that would influence the most appropriate type of cloud including cost, performance, other software inter-dependencies, simplicity, compliance, application certifications etc.

A multi-cloud strategy does not just mean buying cloud services from more than one cloud vendor, it is also mean using multiple types of cloud. Users should use both types of private and public cloud within their portfolio as each has their place in a large environment and users should also consider adopting a combination of different types of cloud i.e. SaaS, IaaS & PaaS to deliver their organisation’s future cloud needs.