The Cloud is all about service provision, but the all important Service Level Agreements (SLAs) used to measure its performance and capabilities still have very little to do with actual service. This is something that really does have to change is service providers are to demonstrate that they can meet customers real needs. The problem is that just about all current SLA models measure the service providers’ infrastructures, while most users want an understandable measure of their own operations.
As Colin Rowland, Senior EMEA VP of OpTier, puts it: “Since the early days internal SLAs have been measured at the infrastructure level – if the CPU is working and the network is up that equals a good SLA. Basically they got written down, put into a filing cabinet and forgotten. Now people want the belief that they can measure something meaningful – that their application is being measured rather than the infrastructure. They want to know that their application is being monitored.”
His argument is that this is where Cloud service providers now need to pitch their service offerings. It maps on to the fact that, for most users, infrastructure technology is now irrelevant. They should not care what server their application or service is run on and, unless it is a financial applications, probably not care which country it is run in. They are now looking for end-to-end service performance measurement defined in terms that are meaningful to them – not the service provider.
This does address a long-standing issue of the taxonomy that should be used in Cloud Services. In Rowland’s view the service providers have been far more concerned about how the infrastructure - the servers, storage and network - have been working rather than what they are working to achieve.
“I have heard talk of a bridge between IT and the business so many times,” he said, “but at least by measuring actual transaction performance the two sides can start to have a dialogue. Business can, for example, say `our users feel the service is too slow’, and the service provider can respond with details about transaction performance in terms that the end users can understand. They can say `trade number 364 took this long to complete’.”
If necessary, they can then optimise the performance or resources committed to a service.
The OpTier solution
OpTier’s solution to this is to measure all the transactions that are occurring in a system, because that way it is possible to identify the transactions and log how long they take to process through the application sequence. That way there is a benefit for the cloud provider in that, if anything goes wrong, they can quickly identify both the issue and the machine. This has the potential to satisfy both users and service providers.
To make the service congruent with user requirements there is a need to identify the most appropriate transactions that need to be measured. OpTier can help customers define these transactions, but this is in practice a post-hoc operation, as the basic service measures all transactions as a matter of course. “Most users will say that all transactions are the most important anyway. And the service is defined by the transactions involved – even logging in to the service is a transaction and should be measured in that way,” Rowland said.
OpTier is installed with the application logic on every machine in the network, and acts like an agent. This automatically picks up any transaction that goes across any machine – virtual or real – and gives that information to the shared service manager. It also reports to the business as well. Each transaction carries an identifier which is scanned and logged at each process step – in much the same way a parcel is scanned at each courier service depot along its journey. These identifiers also contain data used in the transaction classification process.
The company offers two related solutions. The primary one is for all the systems inside the firewall, but there is also a secondary that provides end-user experience management. This sits at the edge of the network, trapping and analysing all the data coming in to it from outside the firewall. This would be such data as browsing or purchase transactions from consumers. This can system can use associated log records to identify how much time it spent on the user’s PC, as well as the time spent on different parts of the network. This is fully integrated with the primary tool, so it is possible to see full end-to-end performance even beyond the firewall.
The first step is to classify the transactions, and then measure their passage the Cloud. This means logging when a transaction leaves a user’s desktop, how long it is on the network, how long it spends on each tier of an application, and how long it waits for the database. That way it becomes possible to draw a complete picture of what happened to that transaction from the moment it is initiated to moment the response or result is returned to the user.
The OpTier service is targeted at both end users and service providers. So far the majority of interest has been from end user businesses setting up their own private cloud services, where shared server farms are running environments like WebSphere, WebLogic or Apache and all applications have to run in that environment. The driver to adopt OpTier has been the provision of an answer to Line of Business managers’ instant response: `what’s the benefit for me and my service requirements?’ The answer is cost and service, and OpTier helps with demonstrating the latter.
But there is now a great deal of scope for Cloud service providers to start offering such capabilities. Its benefits are two-fold: firstly it would allow them to offer customers a value-add, a transaction measurement service that provides them with information, in a meaningful form, which could be of value to the customers themselves and, depending on the actual business, their customers as well. Secondly, it would allow the service provider to measure the actual performance of their own services while at the same time identifying any problem areas quickly. This could be of particular value too SaaS and managed service providers where their responsibility extends to the running of the applications and services as well as the infrastructure.
“We are starting to see service providers move this way,” Rowland said. “They have really good people running infrastructures but they are starting to realise that they have to extend their business models into the services areas and are now starting to look at employing good applications people and tools.”
Post new Comment