While the business case for cloud computing may be clear, strategies for managing service performance are nebulous at best. This article will focus on:
Top cloud monitoring challenges
Key capabilities for tracking cloud performance
Holding cloud vendors accountable for performance problems
Cloud Monitoring Challenges
The primary challenges of ensuring reliable cloud performance revolve around the lack of metrics for monitoring SLAs and performance. Understanding these issues is critical for developing successful monitoring strategies.
1. Obtaining application performance agreements
While vendors will highlight service or resource availability in their SLAs, application performance metrics or expectations are typically absent. In the case of Salesforce.com, SLAs (if one is provided) discuss downtime, but there aren’t any performance guarantees.
2. Lack of performance metrics
Similar to SLAs, organizations should not expect vendors to provide any meaningful performance metrics beyond service and resource availability. If a manager relies on trust.salesforce.com to track CRM service performance and availability, they are limited to monitoring server status, transactions handled, and server processing speed. These reports fail to provide meaningful performance metrics to evaluate service degradation issues or to isolate problems to the cloud vendor.
3. Meaningful Performance Benchmarks
Most SaaS vendors don’t offer any benchmarks or averages that allow you to forecast potential performance or service demand. While IaaS and PaaS vendors will provide cloud performance benchmarks, these numbers don’t take into account Internet latency, the location of your users to the services, or your network environment. The challenge is to properly benchmark and create meaningful metrics for your organization.
Effective Cloud Monitoring Strategies
Given the lack of metrics to assess performance for a specific organization, how do engineers successfully manage user interaction with cloud services? These challenges require organizations to take a holistic approach in monitoring by implementing solutions that allow them to seamlessly view external components and performance as if they were a part of their internal network. Network teams should have a true view of performance that tracks packets from the user over the ISP to the cloud provider, including any cloud-hosted components.
1. Ensuring Service Performance
While SLAs may not guarantee performance, cloud vendors should take action when clear proof shows their systems are the source of the problem. Set up synthetic transactions to execute a specific process on the cloud provider’s site. By regularly conducting these transactions, monitoring routes and response times, you can pinpoint the potential source of delay between the internal network, ISP, and cloud provider. This data along with web error codes can be provided to the cloud vendor to help them resolve issues.
2. Overcoming Lack of Performance Metrics
Depending upon the service, vendors will provide varying levels of detail. The type of service also impacts what you can monitor. In the case of SaaS, you may only be able to monitor user interactions and synthetic transactions, and rely on vendor-provided reports. For Platform as a Service (PaaS), monitoring solutions such as Observer Infrastructure provide significant performance metrics. These metrics can be viewed alongside response time metrics for a more complete picture of service health. In the case of Infrastructure as a Service (IaaS), you have access to the server’s operating system and applications. In addition to polling performance metrics of cloud components, analysis devices can be placed on the cloud server for an end-to-end view of performance.
3. Addressing Performance Benchmarks
With monitoring systems in place, it’s important to baseline response times and cloud component performance. From these baselines, meaningful benchmarks can be set for your specific organization. From this point, alarms can be set to proactively alert teams of degrading performance. Utilizing long-term packet capture, it’s straightforward to investigate performance from the user to the cloud and isolate potential problems.
In adopting monitoring strategies for cloud services, it’s important to establish performance metrics and baselines for your specific organization. This involves going beyond cloud-vendor provider data, ensuring your monitoring tools provide client-to-cloud performance views, and proactively benchmarking and monitoring performance. Taking a holistic approach ensures your staff can manage any performance issue and provide proof to cloud providers if their systems are the cause of an issue.