Containers and serverless: developers are keen but do IT leaders understand them?
Computing speaks to three analytics vendors on what their customers are doing
DevOps is all about delivering better software faster. While they may not be essential to this practice, serverless cloud and particularly container technologies are hoving into view for many companies that are going down the DevOps route.
From a developer's point of view, both allow abstraction of the infrastructure allowing them to focus on the business logic of their code and get it into production quicker. However, both technologies are relatively immature and adoption so far is largest in those sectors where software acts as a differentiator such as media, technology and retail.
However, there are sizeable pockets of users in other areas too: the DWP is a major deployer of Docker and Kubernetes, for example, while Coca Cola is a user of Amazon Lambda, and containers and serverless are to be found in innovation centres within banks, manufacturing and other large businesses.
Of the two, containers are the more established, general purpose way of distancing applications from the underlying infrastructure, and orchestration solutions like Kubernetes offer the promise of making them truly platform-independent; use cases for serverless are currently more narrowly defined. Which is probably why, when asked about their area of focus for a forthcoming project during a recent Computing research analysis, more than twice as many DevOps practitioners said containers than serverless.
We spoke to three vendors who, while not directly engaged in ‘core' DevOps (CI/CD, config management et al), are certainly part of the ecosystem. All have customers that engaged in DevOps. How do they see the technological and cultural battleground panning out?
"These technologies are coming up more and more frequently," said Martin James, regional vice president Northern Europe at big data firm DataStax. "There's a performance improvement that serverless can offer, but taking advantage of that improvement does require more skills around software."
Serverless and container technologies may be relatively raw, but the rate of adoption is rapid says Christian Beedgen, CTO of machine data analytics firm Sumo Logic.
"I think that developers are moving to cloud and to containers pretty quickly," he said. "Based on our data, around a quarter of enterprises in Europe are using Docker as part of their AWS instances. Lambda and Kubernetes are rising sharply in popularity."
Unless you are firmly attached to a particular cloud or enterprise software vendor, it is wise to keep your options open, said James. Platform-agnosticism is the way of the future.
"The most important thing here is not to tie yourself into any one cloud provider as you move to serverless or to containers. You should be able to run across any cloud service, rather than being tied to a specific cloud provider's tools. This kind of data autonomy will be more important in the future."
Innovation centres or ‘internal start-ups' are an excellent way to assess the potential of new technologies, but whatever the strengths of the technologies themselves, how they fare thereafter that will depend very much on how those initial experiments are sold to rest of the organisation, said David Wyatt, vice president EMEA at analytics platform provider Databricks.
"Different teams will adopt these technologies at different rates, and other teams will then look to get involved when benefits become apparent. Increased adoption will be based on how well those initial projects are thought out, designed and results-oriented."
Starting small
The principal of starting small followed by staged expansion applies to DevOps as a whole. A general appreciation of the importance of DevOps and Agile is a prerequisite. This may be meat and drink to web based and ‘digital native' organisations but it's certainly not the case elsewhere.
"DevOps is well understood in the developer community, somewhat understood in the IT operations sector, and not at all outside that group," James said, adding that for organisations that depend on software for differentiation time is of the essence.
"The problem here is that DevOps is going to be essential for companies that want to build new applications that can work in the cloud and at scale. This might seem like an in-the-weeds technology problem, but it is part of how companies will deal with their issues around competing with other companies entering their market. Understanding how you can roll out software faster and help customers in the moment should be on the business agenda."
A common mistake is to simply relabel existing practices as Agile or DevOps without embracing the change that the words imply.
"The big challenge is getting people to actually implement them, rather than sticking with their existing processes and calling them by new terms," Beedgen said.
It may be easier said than done, but organisations should look beyond the hype and the inevitable teething problems to see what the services have to offer.
The understanding around deploying new applications is currently limited
"The understanding around deploying new applications is currently limited - there are few people with the right combination of skills at the senior IT level that can put these new technologies into context for management teams," said Wyatt.
"For the C-level, areas like digital transformation are becoming more understood, but how they get from where they are now to these new IT-led businesses is difficult. DevOps is essential to delivering that, but it does need more simple, business-level examples in practice."