While some applications in the data center require extreme performance, high performance is now a default requirement for all production applications. How performance is measured varies by application (for example, one application might require very high throughput while others might require ultra-low latency). Meeting these varying business-driven service level requirements may require two types of storage architectures: high-performance all-flash arrays (AFAs) that deliver hundreds of thousands of IOPS and low latency, and the emergence of a new breed of extreme performance arrays that deliver one million or more IOPS with extremely low latency. Previously, Storage Switzerland explored the extreme performance use case and architecture. In this blog, we will discuss in more detail high performance use cases, and what an appropriate architecture should look like.
Defining the High-Performance Use Case
High-performance use cases still require large data sets to be processed quickly. However, they are less demanding, when compared to extreme performance use cases, in terms of the volumes of data, their required performance levels, and the consistency of that performance. Typically, these use cases include application consolidation and cloud infrastructure hosting. Server virtualization, virtual desktop infrastructure (VDI), and database management are a few specific examples. Common characteristics are the ability to support increased virtual machine density, multiple users, to scale both up and out, and accelerate application processing performance while simultaneously keeping infrastructure costs in check. Using high-performance storage as the initial backup target should provide better performance to data protection applications with instant recovery features. Re-instantiating a virtual machine on a slow backup appliance with high capacity hard disks may not provide the user with the “instant” response they expect.
Defining the High-Performance Storage System Architecture
Extreme performance systems are typically purpose-built to deliver consistent levels of high performance and extreme low latency for a specific task. However, it is more important, in the case of high-performance storage system architectures, to optimize price and performance while also providing a robust feature set. With this in mind, consolidating the storage footprint (thus lowering power consumption and the amount of hardware required to achieve the same usable capacity) without sacrificing application performance is key. Selectable inline data deduplication, that may be turned on or off by volume, can help to optimize available storage capacity, while also minimizing the deduplication processes’ performance impact on the application. Selectable deduplication is particularly important for workloads, like databases, that will see no benefit from its use. The ability to deploy the system in a dual-controller, active-active configuration better balances application loads, increases throughput and response times while further improving utilization compared to an active-passive deployment. It also facilitates high availability.
Why Most Enterprises Likely Need Extreme and High-Performance Systems
All-flash is quickly becoming the de facto standard for the majority of production workloads. Increasingly, it even makes sense for some secondary workloads as well. That being said, it is important that IT professionals have access to both high-performance and extreme performance systems in a way that most effectively supports business outcomes for as low a total cost of ownership (TCO) as possible. To achieve this end, high-performance and extreme performance arrays should work together. Database environments are a good example. While core database processing may require running on an extreme performance system, a high-performance system may be suitable for test and development and disaster recovery.
In the next blog in this series, we will further explore why the system architecture over the storage media itself matters for both extreme and high-performance systems. Also, register for our webinar “Flash Storage – Deciding Between High Performance and EXTREME Performance” which is available now on demand. In this webinar, we detail high-performance and extreme performance systems, the use cases they address, how to decide which workloads should go on which system type and how IT planners should integrate them.