Reducing public cloud overspend
With the rapid expansion of cloud usage sprawl of resources and optimisation of services becomes an increasingly difficult task. We have already discussed how serverless is on the rise, due in part to greater choice and maturity. This also is combined with the accelerated growth of an average cloud environment.
What this all contributes to is the expenditure of cloud becoming a greater concern. With the adoption of a pay-for-what-you-use model, cost control, analysis and overspend prevention strategies has many cloud consumers concerned that they’re not only “breaking the bank” without realising it, but are also unprepared having come from an on-premises model where this cost issue simply does not exist.
This leaves many customers potentially exposed. Having a strategy to simply review cloud costs isn’t the answer. What’s important is how those costs are analysed, how wastage is detected and what the strategic outcomes. This means having processes and most importantly access to data that can aid in consensus and decision making.
To give an idea to the scale of the problem, most cloud environments have three critical cost areas for analysis.
Wastage
This is where resources are no longer required but still exist in the environment, consuming spend. This is reflective of an on-premises behavioural trait where “clean up” procedures have always been treated as low priority as there is generally no risk or reason to prioritise it over mission critical work. In the cloud however this matters much more due to the cost of consumption.
Performance over-specification
Resources in the cloud have compute power assigned to them for the needs of the workload. However, this is typically very poorly optimised as it requires accurate understanding of what the workload needs to remain performant. In an on-premises environment, IT teams would typically over-spec to ensure performance with very little budgetary consequence. In the cloud however, resources that are over-provisioned are resulting in unnecessary cost to the business with no benefit.
Bad practice
Configuration of resources in the cloud can be done in many different ways depending on requirements, but it’s easy for a configuration that seems to be valid to have an unexpected impact on cost or where best practice has moved on since the original resource was created and things have not been kept up to date. With the cloud, things move at a rapid pace, and what could’ve been a best practice and best cost deployment 12 months ago may not be the case today. For example, virtual machines compute power is defined by a SKU in the cloud, and newer SKUs are added constantly, offering better capability and often, lower cost due to maturity and demand.
Overall, cloud environments can attribute up to 35% of spend to these three areas alone, often creating a negative picture of cloud usage or starving other projects of resource due to budgetary constraints.
This is therefore having a significant knock-on effect, especially where costs are under a microscope in the current economic climate.
To address these areas however is not a simple task. While some corrections will be obvious, how do IT teams optimise for performance?
What is the “correct” performance?
And how can we validate that some resources are indeed classed as “wastage” if no immediate data or evidence is available?
In short, cloud customers need access to a source of truth to resolve this. This is where cloud consumption analysis needs to follow a particular process.
The first part of this process is difficult to master though; how to obtain data, and what data should that be.
One important aspect of this data analysis is that the data should be accurate and up-to-date. The cloud moves fast, so utilising data from a point in time 6 months ago will unlikely be correct and will lead to poor decision making.
There are various ways however to obtain data on cloud resources and access the data for this process using data collection services.
All resources provide a way to stream telemetry style data into a centralised data collection service and cloud services also log all critical actions. Data collection services differ based on requirements, but all generally follow the same principal of making data available to be filtered for easier decision making.
In Microsoft Azure for example there are various data collection services for differing requirements.
Azure Log Analytics is a very popular data collector service for example that allows the streaming and capture of every major Azure resource type.
This then allows customers to review data that is relevant, up-to-date and accurate in one single place to aid in accurate decision making, and more often than not speed up consensus on ways forward to control spend.
Did you know?
IA-Cloud from LABS as a cloud management platform is at the forefront of cloud usage and capability, designed by technical cloud experts who understood the future demand for such services from the outset. As a result, serverless management and monitoring hasn’t been added to IA-Cloud as an afterthought, it has always been a core part of the platform.
IA-Cloud supports serverless in two fundamental ways:
Monitoring – IA-Cloud is unique in the market in that it detects serverless resources, automatically onboards them for management and monitoring and baselines against industry best practice, even down to containerisation services.
Analysis, Optimisation & Cost – IA-Cloud’s embedded eConsultant provides invaluable recommendations on serverless usage, particularly where resources can be better optimised or where best practice is not being followed. In addition, customers can now review recommended changes on how it impacts running costs, as well as the ability to do charge back/show back.
Mining data, applying analytics and surfacing consumable information is critical for reaching a consensus with business leaders on how to address overspend. However, ensuring insights are accurate and actionable takes a significant amount of time and effort, and most importantly expertise in cloud technologies.
With limited technical skills available in the market, this does present a problem.
IA-Cloud from LABS addresses these two critical areas by applying advanced analytics to data on how to correct resources that are overspending or are highly probable as being classed as wastage - with evidence to back up these decisions, for platform services as well.