Cloud Computing in the Network Data Center

Cloud computing is becoming a critical function within the network core as operators strive to improve efficiency, deliver added value, and create new business opportunities. While these trends promise important benefits, they also present challenges. Telecom cloud services must blend data center and telecom capabilities, offering excellent performance, cost, energy, space, and manageability, along with carrier- grade reliability and security.

In this article, we look at the forces driving telecom cloud computing and explore the challenges of adopting this technology. We then examine how Lanner meets these challenges with 2U carrier-class appliances capable of 400 Gbps throughput, and consider how the latest Intel® technologies – including processors, chipsets, Ethernet controllers, and switches – power this solution. Finally, we show how pre-integrated communications software simplifies development and speeds time to market.

The Telecom Cloud Opportunity

With cloud computing transforming nearly every industry, telecom providers have a unique opportunity. Not only can they use cloud computing to improve their existing operations, but they can also take on new roles as cloud providers. With their well-established network- based businesses, local presences, customer relationships, and aggregator expertise, telecom providers hold many advantages over other cloud providers.

The move to software defined networking (SDN) is a key enabler for telecom cloud computing. SDN allows telecom providers to deploy virtualized network functions (VNFs) when and where they are needed to launch new services, create new businesses with ecosystem partners, and improve network operations. Latency-sensitive applications can be hosted locally, while less-sensitive applications can be centralized for cost efficiency. In short, SDN enables the network to be programmed, managed, and optimized more dynamically and at a larger scale.

To succeed in this new context, operators are moving from “pizza boxes” that provide narrow functionality. The cost, complexity, and scalability of these appliances do not suit network data centers. Instead, they need a solution that consolidates functions in as few devices as possible. The Lanner Hybrid Telecommunications Computing Architecture (Lanner HybridTCA*) platform was designed with these needs in mind. Initially launched five years ago and updated with the latest Intel® technology, this innovative carrier-class platform integrates control plane and data plane in a single 2U appliance, delivering up to 400 Gbps throughput. As shown in Figure 1, HybridTCA employs an easy-to-configure modular design.

As the name implies, HybridTCA is based on AdvancedTCA* (ATCA*) technology. What sets the two apart is HybridTCA’s flexibility. AdvancedTCA standards limit each board to 200 W. With HybridTCA, everything is customizable. For example, each board on the HCP-72i2 HybridTCA* Communication Platform can support up to four CPUs with a thermal budget of 130 W for each CPU. HybridTCA is also up to 70 percent smaller and up to 70 percent more energy efficient than traditional appliances. In fact, a 2U HybridTCA design can outperform a 5U ATCA design.

The HybridTCA architecture centers on the middle plane. This board connects the other elements together, making cable connections unnecessary. In back, the Back Blade Unit (BBU) can host two processing blades. Using the latest Intel® Xeon® processor E5-2600 v3 product family, each board can host dual-socket configurations with 24 cores for a total of 48 cores across two BBUs. The interconnection between these two boards goes through the Non-Transparent Bridge (NTB) port in the CPUs, achieving a bandwidth beyond 30 Gbps. Because the architecture can support up to four CPUs per BBU, the system can expand to higher core counts in the future.

The Front Blade Unit (FBU) supports up to three swappable Ethernet interface blades that can be configured with up to 36x Gigabit Ethernet (GbE) or 24x 10 GbE network ports in an array of small form-factor pluggable (SFP) or copper combinations. The front storage unit supports up to three 3.5″ hard disk drives (HDDs) or six 2.5″ HDDs or solid state drives (SSDs). Below the storage area are the USB and serial ports and the intelligent platform management interface (IPMI) management port.

The HybridTCA features 1600 W redundant power units, and all system fans are placed on each BBU for easier service. Through FIPS/NEBS certification, hot plug and hot swap capabilities, redundant power supplies and fans, and rich IPMI manageability, Lanner HybridTCA platforms deliver high levels of reliability, availability, and serviceability.

The modularity makes configuration easy. For example, to create a high-end SDN switch, the platform could integrate a BBU switch based on the Intel® Ethernet Switch FM6000 with a BBU control board based on the Intel Xeon processor E5-2600 v3 product family. For network functions virtualization (NFV), it could consolidate accelerators via FBUs to enhance performance. The platform is also well suited to software-defined storage. To meet the needs of these applications, Lanner created the FX-3710 shown in Figure 2. This 3U rackmount cloud storage appliance integrates two BBUs based on the Intel Xeon processor E5-2600 v3 family with up to 20x 2.5″ SATA/SAS HDD drive bays and six front-facing PCI Express* (PCIe*) slots for impressive storage capacity.

Under the Hood

To make the HCP-72i2 a powerful carrier class-performance platform, Lanner draws on the latest Intel technologies:

  • Intel® Xeon® Processor E5-2600 v3 product family – Based on Intel’s Haswell microarchitecture, these new processors provide larger caches and up to 20 percent more cores than the previous generation. Combined with hardware enhancements like Intel® Advanced Vector Extensions (Intel® AVX) 2.0, these improvements provide up to 2.2x the performance over the previous generation.
  • Intel® Communications Chipset 89xx Series – Designed for communications workloads, this chipset includes Intel® QuickAssist Technology to increase workload efficiency by offloading compute-intensive cryptographic, compression, and packet operations. Lanner supports this chipset in its FBU networking blades.
  • Intel® Ethernet Controller XL710 Series – This next-generation controller delivers proven 10 GbE and 40 GbE connectivity, plus extends Intel® Virtualization technology (Intel® VT) beyond the chipset to deliver industry-leading I/O virtualization innovations and performance. These Ethernet controllers are available in the FBU blades.
  • Intel® Ethernet Switch FM6000 Series – This fully integrated wire speed 10/40 Gbps Ethernet switch silicon provides a practical way to transition to a flow processing-based OpenFlow-type protocol while supporting traditional switching and routing protocols. It supports 10/40 GbE low-latency switching with features like flexible frame header processing for SDN environments. This switch is available on a BBU or FBU.
  • Data Plane Development Kit (DPDK) – This set of libraries and drivers for fast packet processing runs in the Linux* user space environment, assisting in managing memory, buffers, queues, and flow classifications. Lanner experiments show that using DPDK, packet forwarding can achieve line speed even with small 64-byte packets. This performance transforms Intel® processors into an excellent networking architecture. The combination of DPDK and Intel® QuickAssist technology eliminates the need for specialized network processors in network appliances, and enables Lanner HybridTCA to handle all control, data, management, and compute using Intel processors.

Lanner HybridTCA also integrates the Wind River* Intelligent Network Platform. This optimized Linux*-based software builds on DPDK to enable application acceleration, deep packet inspection (DPI), and flow analysis in a single system (Figure 3). By including all the components necessary to consolidate management plane and data plane applications, the platform can save development teams several engineering years building and testing a system. (Like Lanner, Wind River is an Associate member of the Intel® Internet of Things Solutions Alliance.)

Use Cases

The HybridTCA platform can be used in a wide array of data center applications. For example, it could be used as a high availability (HA) appliance. As illustrated in Figure 4, all the Ethernet ports in the front are connected to both boards through a PCIe switch. Each pair of Ethernet ports could be assigned to either of the boards by software. By default, each board controls half of the ports and each board runs its own operating system (OS). Both boards are synced through two 10 GbE SFP+ ports on the management area. If one host fails, the other one will take over and control all the traffic.

The platform can also be used as a DPI machine to provide functionality such as policy control and charging, quality of experience (QoE), and subscriber analytics. The packet handling and compute can be separated onto the two boards, and all the packets coming from one board could be passed to another board through the NTB to achieve the highest efficiency.

A Versatile Solution

Already in use in telecom for application delivery, load balancing, and high-end data center security, the Lanner HybridTCA platform demonstrates the versatility of the latest Intel technology. What’s more, the modular design of the platform can be adapted to a wide range of uses, helping service providers keep up with rapidly changing markets. Blending the best features of data center and telecom infrastructure capabilities, the HybridTCA platform is an excellent match for the needs of the telecom cloud.

Cloud Computing in the Network Data Center was last modified: January 14th, 2020 by LEI Technology