X
Tech

Welcome to the IT service ER

Yesterday afternoon I sat down with Jeff Cobb, SVP, Strategy, at CAworld to talk about the company’s approach to what it calls "service assurance", focusing on how large organisations own and operate IT.As the conversation went on I began to get an image of the modern service desk as the IT equivalent of an ER department at a busy hospital, with teams of staff handling triage, stabilising patients and then passing them on to specialised departments or releasing them.
Written by Simon Bisson, Contributor and  Mary Branscombe, Contributor

Yesterday afternoon I sat down with Jeff Cobb, SVP, Strategy, at CAworld to talk about the company’s approach to what it calls "service assurance", focusing on how large organisations own and operate IT.

As the conversation went on I began to get an image of the modern service desk as the IT equivalent of an ER department at a busy hospital, with teams of staff handling triage, stabilising patients and then passing them on to specialised departments or releasing them. That's the role of a service assurance platform, managing the complex quickly and effectively, pointing service owners to problems, letting them escalate to the appropriate teams – application, network, infrastructure, security, even to third party cloud services.

IT departments are having to become service oriented. It's less a policy they've chosen, more a result of the complexity of the modern IT infrastructure, with mixed stacks of hardware and software. The next stage of this evolution makes things even more complex, as businesses start outsourcing elements of their stack. How do you manage systems when part of a process workflow goes though Salesforce.com or Netsuite, or a key function is hosted in Azure, with data in Database.com? Complexity increases as we move services across clouds and across third parties – making it a much harder problem. There are more variables, more choices, more points that need to be monitored and controlled.

Cobb suggests that organisations should take a model-driven approach to development, deployment and management. You can use models to construct and test applications, examining both behaviour and performance, using black box models of services that aren’t complete or are too expensive to test to synthesise appropriate results. It's another implementation of the design-by-contract model that falls out of service architectures, where known APIs are inviolate links between parts of an application. With a contract all you need to do is get your side of the project right – the rest is up to other teams and partners. It's a dev-ops approach at heart, with a new feedback loop for data that goes to the development teams from operations.

Getting service assurance right is important, as Cobb points out. You can't manage user outcomes by being absolutely sure that one thing is working. Everything might look green, but the end user situation can be wrong as the application semantics might have changed. You need to be able to understand the semantics, and how they affect end user outcomes. You might be meeting a 1ms SLA on a database, but a failure to cache can mean hundreds of thousands of extra calls that end up blocking operations.

Modern management tooling can see what people doing, with sensors and governors in the system, but they're necessary but not sufficient for understanding the semantics. Modern development trends mean that toolkits are high level, and the semantics are in the code. Tools like Java and .NET are run using just-in-time compilers, and management tools can analyse this code, recovering the semantics from them, and then using them to help manage the application and the resulting services.

Cobb talked about CA's data configurable tools, which bring in new modules for new frameworks, adding new modules as new versions of Spring or ASP.NET MVC or whatever roll out. That helps with rapidly changing semantics as frameworks change and update. Alternatively a service management tool can follow the transaction flow in an application, watching information flow at the hub of an application, not at the spokes. That's because it's more important to know what has happened than why, letting service owners triage their own problems – whether they're on premises or in the cloud.

Of course, as Cobb notes, it's easier to manage a service you own. While you might have the same visibility of a cloud service, there are additional issues around policies and legal contracts that are outside the scope of traditional IT management – even though they will need to be part of the process. It's a whole new frontier for IT, bringing it into a closer relationship with the business.

It's an interesting future, and one that's going to make the IT professional increasingly important. And it's also one where watching old episodes of ER might just count as training…

Simon Bisson

Editorial standards