White-box CPEs, generally known as vCPE or uCPE, have become the preferred options for enterprise networks, as the economical values outweigh the traditional proprietary hardware. Therefore, major telecom service providers have announced that they will increase the use of white-box CPE (customer premise equipments).
As addressed by AT&T in recent summit conference, white-box hardware CPE makes more sense for cell sites and central data centers, as they are more flexible and scalable for deployments, saving the time and efforts. On the other hand, 5G is anticipated to be a distributed networking architecture, and thus it is more economical and agile to use white-box CPEs.
However, despite the easy deployments, performance could be an issue if the white-box CPE is not properly configured. Traditionally, the proprietary devices are running on dedicated processor to specialize in required tasks. That’s reason why some mission-critical applications still use proprietary hardware today due to their purpose-built characteristics. On the contrary, white-box CPE are mostly based on Intel® x86 platforms as general purpose computing systems. Without proper configurations, the performance may be obviously frustrating, unless high core count Intel CPUs are used. However, it’s unlikely a branch network would implement a white-box CPE powered by top-tier Intel Xeon® or Core™ CPUs as the cost issues go. Therefore, proper configurations are necessary.
Data/Packet Acceleration Configuration
In today’s SDN and NFV deployments, white-box CPEs are used to run multiple VNFs (virtualized network functions) and create service chaining. Most enterprise IT managements that deploy white-box CPEs would run their VMs (virtual machines) in a Linux OS for its compatibility. Therefore, data packet acceleration is necessary to boost the performance in a Linux environment. A widely used technological configuration for white-box CPEs is Intel DPDK (Data Plane Development Kit).
DPDK is a software set of library and drivers compatible on any Intel x86 processors for accelerated packet processing. It is designed to run in Linux environment for enterprise and industrial network management. The libraries of Intel DPDK improve data plane performance, pre-fetch data, trim memory latency down and reduce development time so that enterprises can save tremendous cost in vertically-integrated and monolithic hardware such as evolved packet cores. In short, Intel DPDK is a software scalable technology that improves virtualization performance without expanding hardware investments.
In a virtualized network orchestration, data encryption/decryption and key compression are highly crucial in the security of data and packet traffic. In fact, cryptographic tasks usually occupy large share of computing and memory resources, and would eventually wear down the CPU capacity.
To accelerate the cryptographic process, enterprise IT management shall choose the white-box CPEs built-in with AES-NI (Advanced Encryption Standard – New Instruction) and hardware-assisted QAT (Quick Assist Technology). These two built-in features would offload the CPUs to improve the performance and security in a virtualization environment.
Another ideal configuration is SR-IOV (Single Root I/O Virtualization). SR-IOV is a virtualization method of existing peripheral to enable scalability and improve performance. With SR-IOV, the VMs can directly connect the PCIe peripherals, and in a networking environment, it’s most likely the NIC.
As addressed above, an economically and technologically optimal white-box CPE requires the configuration of cryptographic process, peripheral and data/packet acceleration. For instance, LUNA-D200 is a recommended choice. The desktop appliance is powered by Intel®’s Atom C3000 Series CPU to offer AES-NI, QAT and DPDK for performance boost and CPU offloading, while the hardware supports NIC module.