CA Technologies’ CTO, John Michelsen, took to the stage at the company’s annual event in Las Vegas this week to tell customers and partners that most companies have a “mess of test environments” and that what they should be doing is transforming their organisations so that there is a quality relationship between development and operations.
Simply put, CA is looking to take advantage of the up and coming buzz word in the IT industry – dev/ops. How does CA propose bringing development and IT operations closer together? Michelsen believes that this can be achieved through service virtualisation, the simulation of live production environments to test new applications. This differs from the traditional approach of testing a newly developed application and then throwing it into a real-life production environment and hoping for the best, so says CA.
Although this theoretically makes sense, is there potential for IT departments to face a management disaster a few years down the line? Of course service virtualisation is sensible, companies would definitely benefit from checking the impact of new applications on services in a simulated environment before propelling that application into a live environment. However, with companies already having had to tackle hardware sprawl in years gone-by, which has since been replaced by unruly virtual machines, is simulated environments for dev/ops the next management headache?
“Dev/ops isn’t something you can buy from us [CA], but it is a transformation that you have to go through yourselves. We have a mess of test environments that don’t really look like production - we have to solve that,” declared Michelsen.
“That’s the fundamental capability we deliver with service virtualisation. As we think about helping you make the dev/ops transition, the first cornerstone to that is service virtualisation. Creating live-like development and test environments that are always at your disposal.”
Michelsen proposed three core components to CA’s approach to a successful transformation to dev/ops – service virtualisation, continuous delivery system, and data mining. He outlined each of these components as follows:
• Service virtualisation – This is at the core of everything and is based on CA’s LISA service virtualisation tool. Instead of creating a test environment for a new order management application before putting it into production, a company would virtually put the new system into a simulated environment to carry out performance testing.
• Continuous delivery system – This component is made possible by CA’s recent acquisition of Nolio, and aims to speed up the promotion of technologies through development environments, through test environments, through to production environments. Michelsen said CA can fully automate this process and reduce deployment and promotion activities for most companies from days and weeks, down to minutes or hours.
• Data mining – This creates the feedback loop that is required in dev/ops to make development more effective. Michelsen said that it was easy to understand what was going on with an application in production when applications were being delivered every other year, but automated mechanisms are now required to bring back that feedback due to the frequency of releases.
Mechelson provided an example of a company developing a new order management application to illustrate CA’s vision for dev/ops in the enterprise.
“An order management application gets a lot of transactions at peak and it has to maintain its scalability and performance SLAs for it to successfully keep running in production. Currently, we would stand up a test environment for performance engineering, with a little bit of performance testing to see how it goes when we deploy it into production,” he said.
“That’s not good enough. What we should do is virtually put the new version of order management in production and then data mine the service assurance products to understand exactly how order management is being used. What’s the load pattern? What does it look like? What are the response times at peak?”
He added: “Underneath order management, what is the ERP system doing? How fast is it responding? We virtually put the new version of order management into production, so that we are certain that it will support the peak load when it arrives.”
Computerworld UK spoke to IDC research manager Spencer Izard at the CA World event, who said that CA’s approach to simulation for test and development is a sensible one, but also warned that this provides an opportunity for a new environment that is hard to manage.
“Virtualisation was seen as great for a long time – companies had hardware sprawl, they were growing out of their data centres, and virtualisation was seen as a fix for this. However, ten years on what it has really done is create a beast that is worse than what you had with the hardware side of things. A lot of organisations now can’t manage their virtualisation sprawl effectively,” said Izard.
“The challenge here is that dev/ops has a critical mass in terms of a marketing term – like cloud, big data – but I think it is in a state that people understand it well enough to mention it enough times but without understanding what it means for them.”
He added: “What that means is that they are going to have to look at not just what technologies can help them do, but how they themselves develop, manage and work with their IT departments.”
Izard also said that if companies want to make dev/ops a mainstream approach to development and testing within their organisation, they need to prove it to the business by tackling an application project that will immediately derive benefits that are visible to the board members.
“The thing with dev/ops is that it is very difficult to say up front what benefits it will bring across the whole of your company. For dev/ops to be a success, what development teams need to do is find the biggest, ugliest software project that they have in the whole of their organisation that provides the most business benefit and apply dev/ops to that,” he said.
“Or find something that provides real value, something that the CIO and the CFO would go: wow. Build your showroom first and go from there. Start relative to the most important thing that can give you the most public success.”
Julian Fish, product owner at Serena Software, a company that specialises in application management, also warned that if organisations assume that buying a set of tools from a vendor will streamline them into a dev/ops approach, they will struggle. Companies need to focus on getting their management principles and processes in place to ensure it is a success.
“Dev/ops is beginning to catch on within enterprises, and so there are now acquisitions taking place with big companies starting to move into providing tool sets to ‘do dev/ops’. I have a couple of problems with this: first is that dev/ops is not about tools, it is about process and collaboration across teams. There is no 'one stack shop' that can help organisations go down a Dev/ops route, and nor should there be,” said Fish.
“Second is that dev/ops is tending to get linked specifically into automation of steps, like release management. Again, this is a big part of dev/ops but not all of it.”
He added: “I think dev/ops will start to grow up from this point - what is important is that we do not lose the focus on results and processes, rather than toolsets and products. If it does get steered down that route, then the real value that dev/ops has to offer will be lost.”
Ultimately CA's tool kit for dev/ops is a useful one, one that will help companies develop applications that are properly tested and fit for their live environments. It will also no doubt, if run properly, cut down on release times and reduce the defects that occur in live production environments. However, it needs to prove to its customers that these simulation environments can also be implemented within defined processes and with the correct tools that ensure that organisations' test environments don't end up in even more of a mess than they were before. Unfortunately, enterprises have too much experience in having to retrospectively deal with sprawl, and the agile nature of dev/ops, combined with the ability to quickly spin up simulation environments, provides potential for this to occur once again. However, if managed properly from the beginning, dev/ops using service virtualisation could drastically improve application development for front line business users.