This article was originally published by +OneX before its merger with iqbusiness. Some references to +OneX remain.
A great benefit of the cloud is the flexibility and democratisation of computing power. However, without education and governance, those with this newfound power could run up unbudgeted bills very quickly. There are many simple options available to businesses to reduce cloud costs, while keeping the flexibility, but as the options grow, so too does the complexity.
Below we outline some of the options companies should consider.
Education is probably the most powerful tool in the company’s arsenal. Regular training on the cloud environment, including modules on costs and economics of cloud, will lead to a workforce skilled and aware of costs and possibly even how to reduce current and future costs.
Adding governance to cloud services may sound counter intuitive to the democratisation initiative, however, this can be done in a way to allow flexibility and even be automated. Giving permissions to only IT departments to provision environments is a simple effective approach. However, another way of keeping costs in check is to create standardised images and pre-authorised approval workflows for automated provisioning and deprovisioning of appropriate environments. This blocks the users from inadvertently provisioning highly unsuitable instances while still allowing them the flexibility to do their work.
Shutting down environments certainly do help in reducing costs, for example auto shut off when idle or even using a development environment only when required and shutting it down there after. There are examples where companies have used corporate calendar scheduling to automate the exact times when an environment is available, automatically provisioning the whole environment just before the scheduled time and thereafter decommissioning the environment at the end of the scheduled time.
If the workload is stable, then changing payment plans to reap savings of a longer-term contract to say 12 or 36 months can provide up to 75% savings. With some cloud providers, there is still flexibility baked into the offerings, for example even though one might have signed an annual contract, there is the option of moving the instances to a different data centre, changing and instance or even an operating system. The more stable the workload and the more one pays upfront the more the savings. Based on the rate that cloud service pricing is falling, and technology changing, for most circumstances, we would recommend taking only up to annual contracts. As an extra note on this topic, some cloud providers even let you sell your long-term contracts to provide that extra level of flexibility.
Not all workloads are created equal, production ERP workloads do not have the same requirements as batch processing, research projects or even building AI models. Changing from pure on demand payments to a more flexible plan based on what free capacity the cloud provider has at the time can lead to saving of up to 90%. This certainly is worth investigating but only if the workloads can be interrupted and restarted without impact. There is the ability to lock in a fixed duration, but the savings reduce significantly. Some cloud providers can show historical rates in different regions and as such one can plan, with a high level of confidence, how to not only keep the workloads running but also reap the highest savings.
Understanding your workload’s compute demands in detail and thereafter choosing the most appropriate instance type, for example compute optimised or memory optimised, can lead to saving from not over provision general configurations. In addition, sizing the instance correctly could also lead to savings from not over provisioning.
Pay attention to your data volumes and storage. The best way to manage this is simply to use intelligent monitoring mechanisms to ensure that the amount of disk space is not over provisioned and that data that is not being frequently used is moved to lower cost infrequent or archive storage services.
Setting up budgets and alerts can be one of the easiest and most insightful ways to control cloud costs. Cloud providers go to great lengths to provide you cost dashboards, alert mechanisms, detailed reporting as well as free training and countless blogs on how to reduce your cloud costs. Some even go so far as to provide forecasting on what the monthly bill is likely to be, giving one the ability to be proactive in managing costs and avoiding month end surprises.
Rearchitecting your infrastructure can turn out to be an effective tool in increasing efficiency and reducing the cost of your infrastructure. Simple architectural changes like auto scaling, consolidation and even conversion from dedicated infrastructure in the cloud to Infrastructure As a Service based services. For example, instead of having dedicated NAT or bastion instances, taking on a managed infrastructure service could increase availability and reduce costs.
Probably the most difficult and time-consuming cost reduction technique is application refactoring. This may be a simple or a major project depending on the application, however, redesigning an application to take full advantage of the cloud could lead the most significant overall benefits. Decoupling, microservices and even serverless techniques can prove extremely powerful in not only reducing current and future costs but also the efficiency of updating and deploying future applications.
As a last point in managing your cloud economics, even though the cloud providers make it more costly to work in multi-cloud environments, specifically with regard to data transfer out of their environment, it is worth considering when the costs allow for it. A once off move, isolated project or even ongoing specific service might be significantly cheaper in another cloud provider’s environment and therefore should be considered if it is technically possible.
Exciting or confusing, a Managed Service Provider (MSP) like +Onex can provide ongoing advice and oversight to expertly guide you in making the most out of the cloud.
