Background
As an increasing number of devices connect to the internet, the significance of the network's edge is intensifying. This growing trend is driving an increased demand for edge computing solutions that provide low latency and high performance, crucial for applications such as artificial intelligence, augmented reality, and virtual reality.
A multi-node Edge Cloud Architecture represents a significant advancement in the realm of network and data processing technology. From enhanced scalability and improved data security to better support for IoT devices and energy efficiency, the development of a multi-node Edge Cloud Architecture is pivotal in meeting the diverse and evolving challenges of today's digital landscape.
Requirements
A leading company focused on software-defined wide area network (SD-WAN) solutions encountered a major hurdle. Their dependence on Amazon Web Services (AWS) for cloud services was becoming overly expensive and less effective for their unique requirements. In pursuit of a more customized solution for their SD-WAN Hub, they came to Lanner and Lanner’s software partner for a solution. The system needed to meet the following hardware requirements.
- Enhanced Performance
The MEC platform aggregates up to 12 (2nd Gen) Intel Xeon Scalable processors, 2 high-speed P4-programmable Tofino switching ASICs, and massive storage with support for precision time protocol (PTP), as an all-in-one modular appliance.
- Faster Connectivity
Lanner’s HTCA-E400 features two protocol-independent switch sleds HLM-E110, both featuring the Intel Tofino P4 programmable ethernet switch ASICs to offer dual channel, 3.2TB switching capability. The programmable switch sled can offload open architecture CPU, accelerating high-priority applications and packet processing for latency-sensitive workload, in addition to supporting carrier-grade IEEE 1588 Precision Timing Protocol (PTPv2), enabling appliances to achieve a higher degree of time synchronization for mission-critical applications deployed in Open RAN, MEC, SRv6, and 5G Edge Clouds.
- Multi-Nodes Computing Power
Lanner’s MEC platforms can support up to five compute sleds, each equipped with a 3rd Gen Intel® Xeon® Scalable Processor. With a total of 200 cores count, PCIe-Gen 4.0 higher bandwidth, and 1.2x memory capacity, HTCA-E400 will deliver unprecedented performance for virtualization workloads, and offering up to 1.94x more secure networking connections to support Open RAN and other Edge Cloud applications.
- High Scalability
Each compute sled on Lanner’s platform can support up to five PCIe*16 sockets for a variety of acceleration cards, including GPU and FPGA cards, to improve high-volume workloads and total application performance. The platform should feature multiple 2.5” SATA storage bays for low-latency communication and NIC module slots compatible with OCP NIC 3.0 cards.
- P4 Programmable Switch
Utilizing Intel architecture and P4-programmable Tofino switching technologies, the platform provides network operators more control and visibility, in addition to flexibility to meet the needs of expanding deployments.
Solution
The multi-node edge cloud solution aimed to not only cut costs but also boost performance and reliability at on-premises and network edge locations. The shift from AWS to a more customized and efficient edge computing solution resulted in decreased operational expenses and enhanced performance, vital for the SD-WAN hub's requirements. The solution offers low latency, high performance, and a simple deployment and management process. Additionally, the modular design of Lanner’s HTCA-E400, coupled with adaptable partner software, offered remarkable scalability and flexibility. This allowed for straightforward adaptation and expansion of operations to accommodate future demands.
Powered by 5x Intel Xeon Scalable Processors, Lanner HTCA-E400 is a 4U, short-depth chassis, hyper-converged edge server specifically designed for Open RAN and MEC infrastructure. With support from Intel Tofino P4 and FPGA/GPU acceleration, the scalable HTCA-E400 edge compute platform features programmable, intelligent switching capability that can offload open architecture CPU and is cost-effective in the long term through optimizing protocol-independent and multi-Tbps networking performance, without being compromised by hardware bottleneck.
This hyper-converged appliance consolidates compute, connectivity, and storage with rich expansion for FPGA/GPU acceleration, specifically designed for RAN Intelligent Controller (RIC), MEC Server, Consolidated Cell Site Router, and Fronthaul Gateway in 5G and the edge.