EMA estimates that enterprise customers with cloud-based deployments are running an average of 20% of their workloads over public and/or private cloud. The problem is that cloud deployments often experience performance issues. Since traditional testing methods do not always catch these issues prior to deployment, optimization has become a requirement.
The Challenges of Cloud Computing
Cloud takes resources that might have been local to the users and moves them outside the corporate firewall and over the public Internet, until, in many cases, these resources are half a continent or more away from the users. This reduces visibility while adding in latency. EMA's recent study, Optimizing the Network for Reliable Application Delivery Across the Cloud found that across all types of cloud — public, private, and hybrid — the majority of deployments suffered from performance issues, even though most participants performed pre-deployment testing.
Despite participants performing these tests and even making adjustments to their networks, such as adding bandwidth or changing the type of connectivity, more than half of the time their cloud deployments still experienced performance issues that impacted end users. The problem here could be that their pre-deployment testing may not account for wireless connectivity, which can come in multiple forms: corporate Wi-Fi, public Internet, or cellular. As Wi-Fi was cited by survey participants at the top connectivity method to reach external cloud services, this is a distinct possibility.
Workloads Matter
Workloads vary by cloud type. Private cloud typically carries more complex workloads including mission-critical custom applications. Public cloud is more likely to carry less critical, more general-purpose traffic, such as office productivity and email. Hybrid tends to be a mix of both.
As private cloud workloads are often more complex than others, these deployments are more likely to experience significant performance issues. Custom applications are often more complex and require access to multiple back-end systems. They also tend to have higher security requirements. All these components can impact response times, and our survey data indicates that this is especially true in the case of private cloud deployments in large enterprises (with more than 10,000 employees) and medium enterprises (with 2,500 to 9,999 employees). The study found that 89% of medium-enterprise respondents experienced performance issues that impacted end users, and 70% of those in large enterprises did as well.
Optimization Choices
There are many ways to tackle these performance issues in the cloud. The study looked at the traditional WAN optimization and application delivery controller (ADC) methods as well as alternative solutions, such as content delivery networks (CDNs) and other types of WAN-optimization solutions. While it was no surprise to see the tried and true methods topping the list across the various cloud types, it was interesting to see how the preferences shifted across various cloud types. The takeway is clear: Hardware-based optimization is giving way to software-based solutions.
It was also interesting to note how these optimization solutions were chosen, as the primary drivers were not feature sets, but rather how "cloud-friendly" the solutions were and whether they were a "good fit" for a particular workload. Both "integration with back-end systems" and budget considerations were also common factors in private cloud optimization choices.
If you plan to deploy to the cloud, make wireless connectivity part of your pre-deployment testing. And assume that despite your best efforts, some form of optimization will be necessary to ensure acceptable performance. Until more of our compute infrastructure shifts to support a cloud-based computing model, optimization technologies will be a requirement, not just window dressing.
Tracy Corbo is Principal Research Analyst at Enterprise Management Associates (EMA).
