beawolf - Fotolia
Among the challenges facing organisations trying to monitor the cloud services they use is how to understand what actually goes on inside the provider’s infrastructure.
Amazon Web Services (AWS) recommends that IT departments should collect monitoring data from all parts of the AWS technology stack used in their application so that a multi-point failure can be debugged more easily.
Areas to monitor include all metrics relating to CPU utilisation, memory, disk, networking and operating system health.
However, the inner workings of infrastructure as a service (IaaS), platform as a service (PaaS) and software as a service (SaaS) are often considered proprietary, which makes it hard to for IT administrators to find the right metrics to monitor cloud applications.
Bola Rotibi, research director at Creative Intellect, said: “It hasn’t been easy to get insight. In the past, companies just wanted to see headline key performance indicators, but now data is the new currency.”
Companies have no control of the cloud services they use, said Rotibi. “You can only see service levels, but it is hard to look at an impact analysis for a business transaction that runs across interacting components in the cloud.”
If something is wrong with how the transaction is being processed by one of the cloud providers’ services, said Rotibi, “the providers could say they are meeting SLAs, but unless you are able to trace every aspect of the application to every touch-point in the cloud, you cannot get a true sense of what is really going on”.
The end-to-end monitoring of cloud applications was among the issues highlighted by Specsavers CIO Phil Pavitt at a recent AppDynamics event in London. To get the right level of insight, Specsavers needed to have application performance management (APM) deployed with its cloud providers’ infrastructure.
“Our partners run AppDynamics,” said Pavitt. The AppDynamics APM agent allows Specsavers to gain access to metrics within its cloud provider's IT infrastructure. For example, the high-street optician retailer stipulated that eClinical Works, a SaaS product that is key to its operations, ran APM, the AppDynamics cloud monitoring software.
AppDynamics’ application performance management (APM) is one of a number of cloud monitoring tools now available on the Azure Marketplace. It enables administrators to troubleshoot performance bottlenecks and optimise the performance of their applications running in Azure environment.
It can be deployed to monitor Azure Cloud Services (PaaS) including web and worker roles, Virtual Machines (IaaS), Remote Service Detection (Microsoft Azure Service Bus), Microsoft Azure Queue, Microsoft Azure Remote Services (Azure Blob), Azure Queue (Microsoft Service Bus), Data Storage and Microsoft Azure Blob Storage.
Three years ago, fashion retailer Asos decided to move from its existing on-premise e-commerce application to a microservices architecture hosted on Microsoft Azure.
Speaking to Computer Weekly at Microsoft’s recent Build developer conference in Seattle, Bob Strudwick,, CTO, Asos, said: “We really bought into the idea of PaaS because we don’t believe there is competitive advantage in the business of commodity compute. We have no responsibility for the underlying infrastructure, so we can focus our resources on writing software.”
But when applications are engineered using a PaaS, there are new ways they can go wrong, said Strudwick. “In the PaaS model, the underlying cause of a problem is more obscure.”
Strudwick said Microsoft was invited to help Asos design its application for Azure. “Through the course of repatriating applications to the cloud, we created a really meaningful relationship with the Azure team and the product labs,” he said.
As such, Microsoft was given a seat at the Asos IT planning table and was in a position to influence architectural decisions. “This manifested itself in architectural reviews and joint performance and resilience test activities,” said Strudwick.
Asos worked with Microsoft to determine what would happen if there were system failures, such as if the data tier in Azure’s North-West region went down. “We modelled the different failure scenarios to understand what would happen to our application,” said Strudwick.
Read more about monitoring cloud applicatiions
- AWS offers an intriguing suite of monitoring tools and capabilities, but many AWS users want a neutral tool that helps reduce lock-in, turning to third-party monitoring services.
- New Azure monitoring tools add another layer of insights into workload performance on the public cloud as providers try to meet enterprise demands and increase transparency.
What became apparent from these stress tests, said Strudwick, was that Asos was not receiving enough information to feed its operations monitoring systems so the retailer could understand how the application was behaving. “We spent a lot of time and effort getting our monitoring and alerting system up to scratch,” he said.
At the time, Asos felt Microsoft’s own monitoring tool, Azure Operational Insights, was not mature enough, said Strudwick, so the company developed its own monitoring based on Elasticsearch and Kibana. These open source tools from logging and analytics company Elastic have helped the company to improve its operational analytics.
Working with the Microsoft product teams, Strudwick said Asos was able to understand how the output data logs from its applications change if there are issues within the Azure PaaS. “It has been a journey for Microsoft, too,” he said.
“For example, the operational processes around the maintenance of the Azure fabric have been made bullet-proof.”