Buyer beware: The ugly truth behind cloud services

The process of selecting a public cloud IaaS provider can be complicated. Evaluating cloud solutions is based on a multitude of variables, but in many cases, comparing them on a level playing field is difficult. During the decision making process, a lot of emphasis is placed on cost, which can be a deceiving metric to determine the true value of a cloud solution.

This recent Cloud Spectator benchmark report provides like-to-like comparisons designed to help IT buyers evaluate virtual machine (VM) performance.

Here are a few use case examples where specific performance metrics were explored around web servers and databases.

Use case 1: Static web server
For web servers such as personal webpages, online portfolios and image galleries, the metrics that are most important are virtual processors and read IOPS. In tests conducted on processor performance and specific CPU-bound tasks, such as data encryption and image compression/decompression, Internap AgileCLOUD’s virtual processors showed around 50% higher performance than comparable Amazon Web Services (AWS) counterparts. From a business perspective, higher read IOPS translates into a faster, more consistent experience for users of static web applications.

Use case 2: High traffic web servers
In addition to virtual processors and read IOPS, high-traffic web servers need to consider write IOPS and network throughput. Examples of these include large news outlets, travel and ecommerce websites that leverage horizontal scalability and serve requests concurrently on a large scale. Poor internal network throughput across a VM cluster can negatively impact traffic distribution across servers, and result in suboptimal application performance.

Use case 3: Clustered database environments
The internal network also plays a major role for clustered database environments, making network throughput extremely important. Other factors such as virtual processors, memory bandwidth and read/write IOPS should be considered in order to achieve optimal price performance, as shown in the chart below.

Decisions about your architecture should ultimately be driven by the performance needs of your application or workload. When looking at specific metrics that matter to your use case, be sure to evaluate the price-for-performance ratio. Many buyers learn the hard way that big-box providers with commodity cloud offerings end up being very expensive in the long run.

Download the report, Performance Analysis: Benchmarking Public Clouds, to see the specifications of the VMs used during testing, along with other database use case comparisons.

Ansley Kilgore