Network Processors and Merchant Silicon Solutions
MicroCode Design and Development for Multi-core Network Processors and Integration of L2/L3 and MPLS Software with Silicon from Broadcom, Marvell, Cavium and others.
Data Plane – The Heart of a Networking Solution
A key aspect of any networking / communications product is the correctness, compactness and efficiency of forwarding in the Data Plane. The performance and success of a Networking Product depends on the performance of the forwarding plane. The forwarding plane in most networking products is implemented in hardware using one of the following methods:
- Merchant Silicon such as Broadcom or Marvel.
- An off the shelf Network Processor such as EzChip, Xelerated (now Marvell) or Cavium.
- In house Proprietary Asic Solutions.
- Processors Designed with Network Processing Cores such as TenSilica.
Each of the above solutions have their own advantages, and the decision that drives the solution of choice largely depends on various factors such as:
- Flexibility and Programmability
- Bandwidth Requirements
- Feature Requirements
Ultimately, the decision is driven by the Use Case and the Market the product is deployed in.
Challenges in Development of Forwarding Plane
The Forwarding Engine whether it be a Merchant Silicon, an Off The Shelf Network Processor or a Proprietary Asic, provides a High Level of Programmability. This is required in order to address a large variety of use cases and also to provide for flexibility in designing unique solutions.
Every solution provides a mechanism to translate decisions made by the control plane into Forwarding Rules that the data plane must apply to every packet. For example,
- Merchant Silicon Vendors publish SDKs that allow developers to program the functionality required.
- Network Processors have an instruction set that must be developed to determine the exact behavior for each packet.
While the fundamental principles for every solution are common, each Silicon Vendor and Network Processor (NPU) has its own unique architecture and programming models.
The Benison team specializes in development of forwarding plane solutions. We understand the critical nature this component plays in the success of your products. We offer complete design and development services for a wide variety of Switching Silicon Merchants and Network Processors.
We also recognize that in order to develop the best solutions, our team not only needs to understand the design of the chipsets or NPUs but we also need to understand the entire networking ecosystem. This implies that our engineers are experts in networking domains such as:
- L3 Routing Protocols (OSPF / BGP / IS-IS / L3 VPN)
- L2 Protocols (Eth OAM / LACP / STP / QinQ and others)
- MPLS and Related Technologies (PWE3 / L2 VPN / L3 VPN / VPLS)
- Multicast Technologies (PIM-SM / PIM-SSM / PIM-DM / IGMP)
- Quality of Service
- Service Plane Applications (Firewall / NAT / JFLOW / IPSec / Lawful Intercept etc)
Expertise in NPU Programming
- Micro Engine Programming
- Packet Buffer Support
- Flow Based and Route Based Packet Forwarding
- Flow Offload Models
- Multiple Pipeline Stages and Interfacing with Accelerators
- Designing TCAM lookup Tables
Experience with following NPUS
Off the Shelf Network Processors such as:
- EzChip NP3 and NP4
- Netlogic (Broadcom)XLR and XLS
- Xelerated (Marvel) X11 and HX-330
- Cavium Octeon Multicore Processors
- Experience with Switching Silicon and SDKs
- Broadcom (Chipset and SDK)
- Marvell (Chipset and SDK)
As an example Benison can assist our customers in the following areas:
- Implement forwarding plane for L2 / L3 / MPLS Switching with linerate performance
- Design a failover mechanism in the forwarding plane to achieve sub 50ms switchover of traffic
- Develop classification methods and queuing models to implement QoS with shaping and policing of traffic