Bivio7000 WP
Bivio7000 WP
Bivio7000 WP
P A P E R
Todays networking applications require specialized network appliances to keep pace with mutligigabit and 10 Gigabit line rates while allowing real-time deep packet processing by the application layer and consolidation of multiple applications on the same system. Neither general-purpose server platforms nor specialized ASIC-based systems provide the performance and flexibility to be the foundation of a scalable, adaptable, and modular hardware platform that provides customers superior cost-performance, investment protection, and flexible upgrade paths. The Bivio 7000 Series of programmable network appliance platforms master these challenges through their unique hardware architecture, Linux-based operating system, open APIs, and system integration support which will quickly catapult the performance of a networking application to the top of the competition without spending the costly development cycles associated with custom ASIC or network processor development.
Figure 1: Bivio Network Appliance Platforms fuse flexible computing with highspeed networking.
Bivio
Control
CPU
Bivio API
CPU
NPU
Packet Processing
By porting their networking applications to the Bivio platform, network equipment manufacturers, independent software vendors, service providers and system integrators can deliver
Market-leading Performance Unmatched Scalability Platform Continuity Broad and Flexible Product Portfolio
Architecture Overview
The Bivio 7000 Series of Network Appliance Platforms is a family of compact, extremely high-performance, and fully programmable network appliances that combine a unique packet processing hardware architecture with a software platform that includes a standard Linux-based execution environment and a comprehensive set of networking features. Designed specifically to provide wire speed deep packet processing, the Bivio 7000 Series architecture fuses Network Processing components with Application Processing CPUs to deliver uncompromising performance and unmatched flexibility. The platform family includes two main product groups that provide performance optimized features to deliver true line rate packet processing from 3 Gbps through 10 Gbps throughput. The major hardware features of the platform include a multi-threaded network packet processor, multiple dual-core application processors, high-capacity Network Interface Modules (NIMs), and a high-performance communication fabric that accommodates full wire-speed data rates between processors. The fabric interconnect also implements a unique stacking technology, which enables linear scaling of application processing power, throughput, and I/O ports via additional processing
1
blades and modules that are housed in the appropriate number of 2 RU chassis. This innovative scaling technology also allows internal failover communication for redundant configurations. The result is a power-packed network appliance in a compact 2 RU base unit with the ability to tailor the computational and network resources to the application profile. System developers can configure their product with 1 to 30 multi-core Application CPUs depending on the application needs. The flexibility of the platform to scale in multiple dimensions allows for optimum matching of various system resources for the desired performance. Furthermore, should application performance requirements grow in the future, the platform enables cost-efficient incremental addition of system resources to relieve any developing bottlenecks without replacing the entire device.
Hardware Acceleration
Application Processor
Network IF
Network IF
In side-by-side performance tests with industry-leading firewall and IDP/IDS applications, the Bivio platform outperformed rack-optimized server appliances by a factor of 10X at 64-byte packets and 3X for large packets. The Bivio 7000 Series of network appliance platforms achieves its breakthrough performance and flexibility by combining a Linux-based environment with three key architectural design elements: 1. High computational power with extensive scalability 2. Strict architectural separation of network and application processing 3. Hardware acceleration Each of these architectural principles will be explored in further detail in the following sections.
Cores can be run independently in asymmetric multi-processing (ASMP) or symmetric multi-processing (SMP) mode. For IPSec VPNs, SSL Acceleration, RegEx content processing and other well-defined computational tasks, a hardware acceleration module can be added to any core through the use of a PCI Express-based expansion module . In a unique approach to scalability, Bivios stacking technology enables application processing to scale linearly with the number of application processors in a stacked system, providing network application developers with the flexibility of tailoring their product to the desired price/performance targets and allowing end-customers to grow incrementally with increasing traffic demands while preserving their initial investment in the system. In contrast to a cluster of independent appliances, the stacked Bivio system configuration is accessed as a single addressable system with one IP address.
Hardware Acceleration
Bivio provides a selection of accelerator module designs to offload a number of computeintensive and repetitive tasks from the application processors in the network appliance platform. These tasks include: IPSec
Each CPU subsystem in the network appliance operates independently of the others, allowing the network services developer to assign different applications to different CPU subsystems and only populate accelerator cards with the CPU subsystems that require offload functionality.
In addition to standard hardware encryption and content acceleration modules that offload encryption tasks and content acceleration functions from the application CPUs, Bivios hardware acceleration options also extend to the programmable packet processor, which provides programmable acceleration of payload traffic that can be controlled by the host application through the Bivio API. For example, an application might identify the service and session of a flow based on the first packet. Next, the application rules are cached and instructions dynamically programmed into the forwarding layer so that all remaining packets in the stream are forwarded in the network layer without consuming further application processing resources or incurring additional latencies.
Inspection Group 1
CPU-1 CPU-3
Inspection Group 2
CPU-4 CPU-10
The default operation of the Bivio network appliance is to load balance all traffic from the network interfaces across all Application Processors. Although this configuration is sufficient for many inline and transparent network applications, the Bivio platform can be easily configured to support advanced network operations in multiple traffic modes.
Inline Tap Mode: In this mode, the platform operates as a transparent inline device while packets are being copied from the wire to the application. Therefore, packets can be sniffed at wire speed and without the need for mirror-ports on a switch. Parallel-Processing with Packet Copy: Sometimes, different applications need to inspect the same packet, but would normally experience resource contention when running on the same processor or shared memory. The Bivio platform avoids these resource contentions by copying packets in hardware to parallel applications without sacrificing throughput or latency. The scalable processing architecture ensures sufficient resources for each application even at full line rate processing. Network-Layer CPU-Offload: In this mode the Application decides which flows to process at the Application Layer and which ones to offload to the programmable Network Layer. This functionality significantly increases the effective capacity of the device. For latency sensitive traffic like VoIP and multimedia, applications can tap into control flows while data flows get forwarded on an accelerated inline path through the network layer, thereby keeping data path latencies at an absolute minimum.
Bivio APIs
Bivio APIs enable system developers to utilize several advanced capabilities of the Bivio architecture as well as offer unique value-added capabilities for custom product differentiation. Bivio APIs include advanced capabilities for traffic modes, system scaling, management and high availability functions on the Bivio network appliance.
Device Management
The Bivio platform supports a command line interface (CLI) as well as a web-based graphical user interface (GUI). The Bivio CLI provides auto-completion, tab completion and context-sensitive help. The Bivio GUI is an easy-to-navigate web-based interface that includes various system monitor and inventory monitor views. Configuration and management of the network appliance platform is typically a combination of calling Bivio-specific CLI commands and application-specific configuration such as policies, users, etc. The Bivio Configuration API can be used to call CLI commands from within a third-party application.
High Availability
The Bivio 7000 Series Network Appliance Platforms provide numerous HA system features that eliminate any single points of failure to deliver non-stop mission-critical services. Standard system HA features include dual redundant hot-swap power supplies, dual redundant hot-swap hard drives with RAID-1, internal redundancy of Application Processors with failure-adaptive load balancing, software- and hardware-based failopen (bypass) network interfaces, and Denial of Service (DoS) attack protection. Additional HA capabilities are also supported including external system redundancy configurations, and an independent management processor with an isolated control network and management port. The Bivio architecture ensures sufficient resources for each application even at full line rate processing.
Bivio 7120
3 Gbps No 12x GE Copper 6x GE Fiber Hot Swap, SATA Yes, AC
Bivio 7130
5 Gbps No 12x GE Copper 6x GE Fiber Hot Swap, SATA Yes, AC
Bivio 7510
5 Gbps Yes 12x GE Copper 6x GE Fiber 2x 10GE Fiber 4x 10GE Fiber Hot Swap, SAS Yes, AC, DC
Bivio 7560
10 Gbps Yes 12x GE Copper 6x GE Fiber 2x 10GE Fiber 4x 10GE Fiber Hot Swap, SAS Yes, AC, DC
Summary
The Bivio 7000 Series of 10 Gigabit Programmable Network Appliance Platforms enable customers to achieve disruptive improvements in deep packet processing performance, scaling, reliability and deterministic traffic management capabilities that network equipment manufacturers, inde-pendent software developers, service providers, enterprises, and federal agencies demand in their emerging networking infrastructure. Network Equipment Manufacturers, Integrated System Vendors and Federal GOTS/COTS Systems Integrators are now enabled to rapidly develop innovative, market-leading networking solutions, speeding time-to-market and time-to-solution with minimal integration effort.
2007 Bivio Networks, Inc. All rights reserved. The Bivio logo, BiviOS, Bivio 7000 Series, Bivio 7100, and Bivio 7500 are trademarks or registered trademarks of Bivio Networks, Inc. All other company and product names may be trademarks of their respective owners. Bivio Networks may make changes to specifications and product descriptions at any time, without notice.