As enterprises increasingly shift to the cloud and introduce multiple cloud vendors, cost is becoming a significant concern. Cloud sprawl - unmaintained orphaned applications and instances left always-on are the equivalent of leaving the car running while parked up.
In this article, we will take a look at some common best practises and why we think cost management is going wrong!
A recent Flexera survey report reveals that cost optimisation was a major focal point for enterprises during 2019, a trend set to continue. Cost is becoming an increasingly worry due, in part, to the introduction of multiple cloud vendors with 84% of enterprises questioned using more than four cloud vendor platforms. Half the audience surveyed state their enterprise was spending more than $1.2 million on public cloud services per year and 23% spending over $2.4 million annually. Amazon Web Services even recently dedicated a specific help hub dedicated to reducing spend with them!
Enterprises are addressing this by establishing cloud expert centres, dedicated teams of cloud experts charged with, amongst other tasks, reviewing cloud costs. 66% of enterprises already have expert centres in place with a further 21% actively considering introducing this as a new business function.
For the most part, these expert centres will likely introduce some common practice solutions.
Public cloud vendors tend to offer computing rates relative to the region in which cloud resources are being utilised. Ultimately, the cost of running the data centre is reflected in the pricing. For example, in areas where electricity is cheaper to produce, and IT skills are readily available, the overall cost for the data centre will be lower, which is passed on to the customer. The popularity of the region, with more customers placing an increased burden on finite resources, is also a cost variable.
The most obvious way to reduce cost is to use alternative regions and making a trade-off with cost savings and overall experience - perhaps accepting the slightly higher response times that come with being physically further from customers, for the costs it save. For example, realtime Microsoft Azure cost analysis indicates an average hourly price difference between $0.54 and up to $1.09 depending on the region.
For the most part, these expert centres will likely introduce and apply the following solutions.
A typical project might consist of multiple instances hosting different environment types - such as development, staging, testing and live environments. Amplified across projects and teams and even multiple cloud vendors, this leads to potentially many instances up and running and consuming resources. While these may be essential to daily work cycles, left unattended, these can quickly cause costs to spiral uncontrollably. The last thing development teams want is their working environment disrupted or inaccessible, so they tend to go unchecked. An expert centre will then typically try to enforce rules on usage times - spinning down under-used environments after hours and weekends.
Generally, computing requirements will lean on the side of over-optimistic, reflected in the computing resources allocated to a project. It is always preferable to consider the "what if" scenario and ensure there is sufficient capacity available to handle any unpredictable usage demands. Applied across multiple instances, this leads to overpowered environments that might only be using a fraction of the resources allocated to them. The expert centre will take a subjective overview of usage and can then start trimming back accordingly to reflect the demands on the environment.
Cloud vendors need to juggle their capacity against demand - they need to ensure they can meet demand with resources to spare, while not being overburdened with too much unused capacity, the cloud equivalent of having too much stock in the warehouse.
In order to ease this, they encourage customers to reserve instances up-front with the lure of more attractive pricing. This can lead to very appealing discounts, with, for example, Azure offering cost saving of around 46% for a year for specific instances types.
An expert centre will try to gauge the requirements of a project and implement an up-front price over a long period. It is an educated gamble between overpaying for something that might not be fully used and choosing PAYG pricing, which will cost more but will be more dynamic and forgiving.
Securing a long-term enterprise contract enables cloud vendors to offer some heavy discounts. As with instance planning, knowing that a customer has a long-term commitment and minimum spend presents some fixed variables for the cloud vendor which can pass on savings. This can be appealing for a expert centre since they can staff-up and train accordingly, knowing a cloud vendor will be a long-term part of their infrastructure planning. The trade-off is actively agreeing to vendor lock-in, where regardless of a changing business environment or needs, the enterprise will be committed to spending a minimum significant amount over many years.
A Better Way
All of these are sound and common-sense suggestions that can lead to tangible savings. However, they are all dependent upon having experts readily available to implementation cost optimisations on an on-going basis. Staying current with new cloud technologies and knowing how to apply them to deliver savings is a daunting task - between them, AWS, Azure & GCP offer over 500,000 items in their product and feature catalogues.
Divio takes a different approach by building intelligence into the platform - a cloud management platform should be managing costs in the background with no manual tweaking or intervention.
If environments become idle or under-used, the resources are optimised in the background. Regional costs are weighed up against project needs, and a best-fit is determined, balancing cost against usage.
All this takes place in the background, all of the time.
Book a jargon-free demo to see how Divio could manage and optimise your clouds.