Why Product Architecture Matters

May 26, 2021

Last month, Array joined the Intel Network Builders program, which is working to accelerate the transition to Software-Defined Networking (SDN) and Network Functions Virtualization (NFV). As of this writing, the ecosystem has more than 150 vendor/members, as well as a growing list of end-user organizations.

Joining the Intel program is another important milestone in Array’s commitment to NFV. (As you may recall, late last year Array joined the OPNFV Project as a silver founding member.) It also serves to reemphasize why our commitment to a CPU-based architecture – vs. relying on ASICs – makes a very big difference in our application delivery controller and secure access gateway products, and thus benefits our partners and customers as well.

ASICs came to the fore some years ago, when the general-purpose CPUs and operating systems of that time were unable to provide the performance required to process ever-growing network traffic loads. As in any design choice, however, there are (and remain) drawbacks to an ASICs-based approach. ASICs carry higher engineering costs, which typically translate into higher product cost, and software bugs can be very complex to fix. Those two factors combine to result in a longer time to market for new features, new capabilities, and bug fixes.

In the meantime, Array developed its SpeedCore® operating system, a next-generation software architecture that allows Array products to take advantage of CPU advances, and to easily scale to meet the needs of complex and high-performance application delivery networking environments.

SpeedCore’s multi-core technology allows Array to leverage general-purpose processors to provide equal or better performance than ASIC-based architectures, with better agility and much lower costs. SpeedCore’s CPU-based environment allows Array to introduce new features and enhancements quickly, and without requiring our customers to rip-and-replace their existing products, or even to take products offline for a hardware upgrade. Instead, Array customers can add features or special customizations with just a simple, non-disruptive software upgrade.

In addition, leveraging the SpeedCore operating system and a CPU-based architecture for our dedicated and virtualized appliances minimizes overall complexity and maintains guaranteed high performance and reliability, while keeping the lid on support costs.

Array’s early choices on architectural design also hold important ramifications for the transition to NFV. The efficiency and agility of SpeedCore and a CPU-based architecture will allow Array ADC and SSL VPN products to more easily accommodate the new NFV model. In addition, Array ADCs have highly granular visibility into applications, allowing them to gather application-level insights that can be leveraged to guide SDN-based switch packets, thus improving performance and security.

Lastly, I/O is one the main performance bottlenecks when virtualizing the network functions. Intel’s ability to support SR-IOV on multiple platforms/hypervisors becomes a key factor in ensuring that Array’s virtualized network functions perform at the highest possible network throughput. Array’s virtualized platform (the AVX Series) uses Intel’s NIC and SR-IOV technology to achieve industry-first guaranteed-per-instance performance on a multi-tenant platform.

As you can see, architectural decisions made early in a product’s life cycle can have a huge impact on performance, agility and reliability far down the line. Array’s foresight in choosing an Intel-based architecture, coupled with our innovative SpeedCore OS, has given users of Array’s ADC and SSL VPN products a wealth of benefits.

To read more about ASIC-based vs. CPU-based architectures, see Array’s SpeedCore white paper. To learn more about SpeedCore, visit our Web site.

Thomas Ching