• Despite the name, there are servers involved, you just never have to worry about them
• They are ideally suited to simple functions with clear use cases that can have a volatile demand such as with IoT devices
• Costs are more transparent as most of the associated costs would be against cost to server (RevEx)
• It is relatively immature technology, but maturing fast
• It separates the users (developers, business, customers) from the infrastructure
Serverless is a relatively immature technology, but it’s maturing fast. Despite its name, serverless does actually require servers. The rationale behind the naming is that your organisation no longer needs to worry about server management, the costs associated or the technologies underpinning it. These are all managed and handled by the supplier (such as Amazon, Microsoft, IBM and Google). Also, as an organisation, you are able to pass responsibility and accountability for the availability of various functions to that supplier, which typically operate at 6-sigma level.
What are the benefits?
In circumstances where serverless is suitable (see below), there are substantial cost savings to be made and increases in availability.
For CXOs looking at budget lines, and particularly those doing zero-based accounting, instead of having to invest large sums upfront in infrastructure, it’s far easier to get the business case through as you only have one Capex cost – which is the project and development cost.
Once up and running, serverless operates on a Pay As You Go basis and potential running costs are lower as serverless executes code in a very precise way, meaning that you will only get charged when the code is accessed and executed, and these costs can be very low per call (see the Coca-Cola case study below).
Cost transparency is also a key benefit. Serverless suppliers can tell the technology business that a particular function is going to cost Y pence to execute, so transparency is driven by the business understanding how often the function is executed. A further benefit for organisations is that serverless is function-orientated, and as a consequence it is more focused and aligns better with back-end applications and IoT functionality.
When should organisations consider using serverless?
1. One area that particularly suits serverless is IoT, because there is limited requirement for a front-end development and simple pathways. Organisations that have started using IoT as part of production or for gathering data for informed decision-making may well benefit from investigating serverless to reduce infrastructure requirements and management costs.
2. Serverless is suitable where you’ve got highly unpredictable supply and demands on your infrastructure. For example, you have a service that has occasional surges in demand that are large (E.g. goes up by 20x) and unpredictable. In order to meet that demand without serverless, you’d need to invest in 200 servers, which would be expensive and difficult to manage. But if that function is hosted on serverless, this surge in demand is easily absorbed by the supplier.
3. If you’ve got specific functions with straightforward pathways, such as getting a bank balance, returning the temperature of an industrial boiler, footfall to your store and loyalty cards these could all be serverless.
4. Where your end users are using voice control. For example, if you are an energy company and want to give consumers their energy use by voice assistant, you have no need to build your own additional infrastructure as you’re creating your own serverless function that can respond to smart assistant/voice requests, such as Alex or Siri.
It’s not the panacea
As the focus of serverless is function points, there are many instances where this may not be the right solution. If you have a large number of functions that are needed to run a service with many interdependencies, the complexity of the solution grows and with hard limits set on total number of functions set by some providers (to prevent DDoS) it’s not the solution to all technical challenges.
There is also a potential lock- in with one supplier as tools that allow you to migrate between providers are not mature. While availability with these providers is typically getting towards 6 sigma level (99.99966% availability) latency can be an issue if you have geographically dispersed data or multi levelled architecture that relies on a hybridised architecture.
We recommend that organisations investigate serverless as part of their overall technology mix. For example, there’ll still be a requirement for most organisations to have some form of infrastructure and its only suited to certain user cases, some of which are outlined below.
Serverless is a demonstration of the continued evolution of options available to help organisations reduce complexity and costs and to deliver functionality. The Coca-Cola example shows just how serverless paired with IoT type devices can provide low cost to serve and high availability for systems that are critical to organisations’ evolving needs while minimising complexity. Serverless is another tool to be aware of, so that when business and customer challenges present themselves, you can make the right decisions.
A showcase of serverless technology is an implementation by Coca-Cola North America. They wanted to have a lightweight way to process payments and increase loyalty from their vending machines across the continent. This implementation comprised a payments gateway, the Amazon API gateway and Apple/Google Pay with push notifications.
From a financial perspective by moving from Amazon Web Service to Lambda (Amazon’s serverless offering) the running costs were reduced by 65% from c$13k to c$5k per annum on purely the usage basis. There were additional cost savings in not having to manage their virtual instances and other hidden operational costs (such as networking, security and enhanced availability). This was for a service with around 1 million calls per month.
Augmented with this is the ability to integrate with existing environments in Coca-Cola and support for continuous integration and continuous deployment. With this comes the ability to continue to work in agile ways and an increased operating cadence as infrastructure teams do not need to be engaged.
You can view the Coca-Cola case study here.