Traditionally, Application Performance Management (APM) has been a bolt-on that comes along after an application has been installed into production. Making sure it is working OK has always been, if not an after-thought, then a post-hoc part of the long-term plan.
So, including APM as not just a part of new applications implementation but also as a core part of the pre-planning and initial design stages for new services marks a new stage in how this fast-developing technology can get used.
Virgin Money, for example, is now starting to work with Dynatrace to achieve exactly this. The company has already installed Dynatrace on all of its online systems to monitor and manage existing and new applications across its internal business-critical applications, such as hosting.net, the Oracle and Java sides, as well as the jBoss and Linux systems.
It is with the new applications that things get more interesting, for Dynatrace is claiming that it has enough track record on products, applications, tools, how they operate and how they inter-operate, that users could get to the point where you start using it as part of the design process for a new service.
In essence, it will be able to inform developers, in advance, that 'Application A will not talk to Application B in that situation'. Andrew Lofthouse, Virgin Money’s Senior IT Analyst responsible for eCommerce and Middleware Infrastructure, says:
Exactly, yes. There’s a bigger discussion on how we move forward, but ultimately at the minute what we’re trying to do is categorise systems as they come into play, into categories of gold, platinum, silver. It’s not necessarily going to be the solution for everything but, more towards the critical business applications we’ll be using it.
Of reliability and more
In the first two days of installing and starting to use Dynatrace, Virgin started to get a full picture of the systems they had installed it on, including the inter-dependencies between the processes and services. In that time, they were getting everything needed from an application and process perspective. Lofthouse recalls:
We also installed it within pre-production environments for load tests and stuff like that, to figure out issues before it goes live, so we know what we’re getting out of it for the pre-production environment before it goes live.
In the end, however, Lofthouse is not solely focused on improving reliability. He is also interested in getting better insight into the existing systems and the scope it can provide to gain time for other work because the team is taking less time investigating issues. In addition, the ability to drill down into tasks that are causing problems is, he finds, speeding up finding the right solutions to operational problems:
Recently we had one where we had hundreds of users within the business failing to log in to the system and the applications. It was a table on a database and the queries were hitting it and building up. It was causing a stuck session. This could have potentially taken us three days with different teams investigating, but we drilled in and within thirty minutes I’d spotted the issue, handed it to the developers, and they fixed it. It was nothing complex. However, if you haven’t got the tools to find out what it is, you’re looking around the estate on numerous hosts, looking at log files; it all that takes time.
This problem can get worse in a load-balanced environment where the error could be hitting any one of the servers. Using Dynatrace, he indicates, this complication no longer existed as the tools identify the specific host involved.
Into DevOps and `reconstructions’
When it comes to expanding Virgin’s use of DevOps, one of the key roles Lofthouse sees for Dynatrace is thoroughly load testing new applications and services before they go into the production environment. And because it is a cloud-delivered service, this allows Virgin to get better value from its investment by judicious re-use of the Dynatrace agents:
We’ve got the agents installed on a lot of our development and pre-production environments, so we load test against those environments to see what things are going wrong. Obviously, we pay for the agents, so we don’t freely put it on everything. But the beauty it does have is that you’re able to disable one of these environments from the dashboard so I can assign it to a different server. I can then install it on that server and say that’s taken that licence place.
So we can transfer these agents around our difference development environments when it comes to load testing. Because what's the point in having one sitting on a development box for a year when the code might only be delivered two or three times? It’s a waste. So let’s take it off and put it somewhere more useful.
The recent Dynatrace Perform conference was the first time Lofthouse had come across information on the new Dynatrace tool, Session Replay, which will allow users to take the logs of a user or customer session and reconstruct it on-screen. In this way, developers will be able to witness any problem that a user or customer was reporting and trying to describe, in the way that they witnessed it.
It is too early yet to say whether it will form part of the Virgin Money portfolio, but he acknowledged that there are parts of the company’s business – especially in resolving any customer-facing issues – where would seem to have obvious potential.
It also has potential as a tool for ensuring that regulations on operational compliance are adhered to. This is obviously an important factor for any business in the financial services sector, where ensuring that transactions are performed in the correct manner can be a crucial factor in managing the business. Lofthouse elaborates:
That is a different conversation and one I’ve yet to have. Obviously we’re a bank, financials, so there’s a lot of PCI compliance issues around customer data and all of that, so how they are going to deal with that I’m not quite sure. Initially it looks great, but there are a few questions that, as a bank, we’d need to ask about before we implemented it.
Assuming that such problems can be resolved, however, he did agree that a system which could monitor transactions for regulatory compliance of transaction processes could create a service that meant there was one less problem to keep track of - and react to post hoc:
Oh yes, that would be very helpful, in more ways than one.
This is a good example of where Dynatrace seems to be heading with its customers. Virgin Money is taking its comprehensive APM tools and starting to explore other opportunities where it can add additional value and help manage a much wider range of business operations.