Over the years, we have learned, designed, and tested several cost-saving measures and techniques that we apply every day to save our clients and customers money, in some cases up to 70%! If you want to know more, keep reading.
Nowadays, opportunities for improvement are everywhere, and removing bad practices is a great way of achieving that. Bad practices are inefficient, generate bad experiences for our clients and customers, and lead to unnecessary security risks.
Today, we want to discuss how to avoid high costs and improve management to achieve significant savings using cloud technologies. Our goal is to help you identify best practices and pinpoint areas for improvement, ultimately bridging the gap between excessive spending and significant cloud cost savings.
Three common money pits
Pit #1: Networking
Inadequate networking designs lead to substantial cost overruns. For example, suboptimal routing between regions or availability zones can increase latency and data transfer costs. Furthermore, inefficient traffic routing between resources can increase cloud expenditures.
Pit #2: Idle Instances
Running underutilized or unnecessary instances is a major contributor to unnecessary cloud costs. Failing to adjust instance sizes to match actual workload demands, often called "rightsizing," results in paying for idle capacity. Additionally, forgotten instances left running without proper automation serve no purpose, further increasing costs.
Pit #3: Unnecessary Capacity
Unnecessary capacity can appear in several forms, the three main ones are:
Instances bigger than required today
Contrary to what happens when managing our own data centers, cloud providers allow us to upgrade our servers instantly. This allows us to not to have to pay in advance for excess capacity in case we need it in the future.
Poorly chosen workload-specific instance type
Related to the previous problem but not quite the same, being wise about the type of instance will maximize the performance for each specific workload we get for the money we are spending.
Let's give an example to illustrate: if you want to run a network load balancer you don't need a huge amount of the fastest storage available so, paying extra for that feature is a complete waste of money that could be used elsewhere.
Poorly tuned scaling
A poorly configured autoscaling could potentially spawns more instances than required and keeps them alive for longer than needed and on top of that, spawns the wrong kind of instance for the given workload compounding the previous two problems and multiplying its effects.
Addressing money pits
Networking pit
This could be one of if not the toughest ones to fix, because it could implicate modifying other systems and/or software so, as always reality, security and careful evaluation should take precedence over everything.
Let's look at some strategies to have more efficient and cheap networking:
Efficient Routing
Design your network architecture to minimize data transfer between regions and availability zones.
In more than one occasion, we re-routed connections to a managed database from a public endpoint (which generates NAT Gateway traffic and internet traffic) through a VPC, reducing the monthly billing by more than 10%; this cost compounded tens of thousands of dollars every year.
Network Security Groups (NSGs)
Most people only consider security groups exclusively as a security tool, but they also prevent unintended and/or unnecessary traffic flows from causing runaway networking costs by means of implementing granular network security rules.
Avoiding public network routing through VPC peering
When data is transferred over the public internet, it incurs data transfer fees increasing significantly cloud costs, especially for large-scale data transfers or high-traffic applications.
VPC Endpoints (AWS only)
When communicating from a VPC to a managed service like S3 or ECR check for Cloud configurations that reduces the costs of data transfers between a VPC and those services.
Content Delivery Networks (CDNs)
CDNs are a powerful but a potentially expensive tool, for some applications can be a money saver but for others they are a money pit; so careful consideration of when deploying them is crucial.
One popular use for a CDNs is to cache content near where its consumed, this is especially useful for global presence. This reduces latency, the size of the managed infrastructure, and also saving on bandwidth costs.
Instance-related pits (Idle instances and unnecessary capacity)
The key to closing these money pits its information; without it you won't know what's happening and much less fix it, after this, the first one is quite obvious.
An appropriate observability tool
You might say: This will increase my expenditure, I have to implement one; and at first glance, might seem like that's right, but as we all know: It takes money to make money; and having the proper read on how your resources are behaving is the key to know what to change to save money.
With all the information from competent observability too you will know for example where your infrastructure is idling and when and as your size increases it comes to a point where you'll be leaking "small" amounts of money everywhere that amount to a LOT; also you will know where and how apply what we are telling you next.
Rightsizing
With proper observability tooling you can continuously monitor your instances and analyzing their resource utilization, this key information will allow you to adjust them with extreme precision to accommodate exactly your workload and nothing more saving those dollars to be used elsewhere.
Spot Instances
Leverage spot instances for non-critical workloads that can withstand being shutdown with no notice will enable you to save costs big time.
Fine-tune autoscaling
Autoscaling is powerful, but sensitive; we should be clear: you can save a lot of money but create a lot of problems for yourself if you are not sure of what you are doing.
Controlling how resources are created and destroyed is somewhat of an art, but when done right, the rewards as high.
In several instances we achieved a 60% cost reduction by using proper auto-scaling and rightsizing. Also, by leveraging spot instances, we reduced 70% of the computing cost without reducing capacity.
Leveraging knowing your needs
Reserved Instances
Commit to using specific instance types for a fixed term to receive significant discounts. This is ideal for steady-state workloads where you know your resource needs in advance.
Savings Plans
Pay a low hourly rate for a consistent usage commitment. Savings Plans are more flexible than Reserved Instances and can be a good option for variable workloads.
Insider information
AWS has a marketplace for reserved instances and savings plans where you can reduce your commitment risk by buying and selling other users' contracts. This is particularly attractive since tools exist to automate this process based on your usage.
Leveraging both plans and when using AWS its marketplace, can lead to 50% savings on computing expenditure. Propper planning and accurate information is a must since you only increase your commitment.
Other cost-saving tools and strategies
Cloud Cost Management Tools
CloudHealth, CloudCheckr, or AWS Cost Explorer help you to analyze and optimize your cloud usage and expenditure.
Resource-tagging
Tag your resources with appropriate labels helps you track costs and identify optimization opportunities with ease.
Promoting a Cost-Conscious Culture
Last but not least (quite the opposite), invest in your team's education, this is a long-term strategy that pay huge dividends across all your teams
Here you have 3 action points you can start to strategize in:
- Educate Your Team: Train your team on cloud cost optimization best practices.
- Set Clear Cost Goals: Establish clear cost targets and track progress regularly.
- Encourage Cost-Conscious Behavior: Promote a culture of cost-awareness among all employees.
Ready to save a lot of money?
By implementing these strategies, you'll significantly reduce your cloud spending and squeeze every value drop out of your cloud investment.
Remember, a well-optimized cloud environment is a lean, mean, cost-saving machine that is efficient, secure, and ready to drive your success.
Still feeling a little lost in the battle against cloud overspending?
No worries! We're here to help. Email us at [email protected], and we can be your trusted cloud, cost-conquering sidekick.
Credits
Writer: Ben Rodríguez
Editor: Luis Vinay
Technical reviewer: João Moura
Researcher: Ben Rodríguez
Illustrator: Dai Fiorenza
Disclaimer:
In this article, AI was used to check grammar and syntax.