skip to Main Content

Napatech Link Capture Software

for Intel® Programmable Acceleration Card with Intel® Arria® 10 GX FPGA

Feature Overview

100 PERCENT CAPTURE & REPLAY
Napatech Link Capture Software turns the Intel® Programmable Acceleration Card with Intel® Arria® 10 GX FPGA (Intel® PAC A10 GX) into a powerful SmartNIC for packet capture and replay applications.

The intelligent feature set offloads processing and analysis of Ethernet data from application software while ensuring optimal use of the standard server’s resources leading to effective application acceleration.

FEATURE

Full Line-Rate Packet Capture
• Zero-loss packet capture for all frame sizes

Full Line-Rate Packet Transmit
• Line-rate packet transmit for all frame sizes

Optimized Packet Buffers
• Low CPU load on the host server

Onboard Packet Buffering
• Ensures packet delivery during congestion

HW Time-Stamping
• 1 ns time stamp resolution

Time Synchronization
• OS time

Frame Processing
• Multi-port and multi-card data merge
• Frame classification
• Frame and protocol information
• Fixed, dynamic and conditional slicing
• Ethernet FCS verification

Tunneling Support
• GTP, IP-in-IP, GRE, NVGRE and VxLAN tunneling support
• Inner tunnel slicing and filtering of packet headers and payloads

IP Fragment Handling
• Identification and distribution of IP fragments

Packet and Flow Filtering
• Configurable filters
• Port and protocol filters
• Pattern compare filters
• IP group match filters with up to 36000 IP addresses
• Up to 36000 flow filters
• Packet discard, packet forward to a specific stream, packet coloring and packet slicing controlled per filter or flow by the application

Flow Identification
• Flow identification based on hash keys
• Multiple protocol-specific hash keys
• Custom hash keys
• Dynamic hash key selection
• Correlation key

Intelligent Multi-CPU Distribution
• Configurable flow distribution over 64 Rx streams

Host-Based Transmission
• 128 Tx streams • Transmit on time stamp

Local Retransmit
• Local retransmit (port to port) with embedded time stamp

Advanced Statistics
• RMON1 port statistics
• Advanced statistics

Monitoring Sensors
• Temperature sensors

data-sheet-feature-overwiev-fig-1

The Napatech Link Capture Software decodes all frames regardless of encapsulations and can identify flows based on user-defined header information.

FULL LINE-RATE PACKET CAPTURE
The Intel® PAC A10 GX with Napatech Link Capture Software is highly optimized to capture network traffic at full line rate, with almost no CPU load on the host server, for all frame sizes. Zero-loss packet capture is critical for applications that need to analyze all the network traffic. If anything needs to be discarded, it is a matter of choice by the application, not a limitation of the Intel® PAC A10 GX.

OPTIMIZED PACKET BUFFERS
Napatech Link Capture Software uses a buffering strategy that allocates a number of large memory buffers where as many packets as possible are placed back-to-back in each buffer. Using this implementation, only the first access to a packet in the buffer is affected by the access time to external memory. Due to cache pre-fetch, the subsequent packets are already in the level 1 cache before the CPU needs them. As hundreds or even thousands of packets can be placed in a buffer, a very high CPU cache performance can be achieved leading to application acceleration. The flexible server buffer structure supported by Napatech LinkTM Capture Software can be optimized for different application requirements:

• Applications needing short latency can have frames delivered in small chunks, optionally with a fixed maximum latency.
• Applications without latency requirements can have data delivered in large chunks, providing more efficient server CPU processing of the data.

The customer can configure the size of the host buffers from 16 MB to 1 TB.

ONBOARD PACKET BUFFERING
Onboard Intel® PAC A10 GX 4 GB memory buffer allows packets to be buffered during micro-burst or PCI Express bus congestion situations.

TIME STAMP
High-precision time stamps, sampled with 1 nanosecond resolution, are applied to all frames received by the Intel® PAC A10 GX. Support is provided for 4 different 64-bit timestamping formats:

• 2 native UNIX formats with 1 ns or 10 ns resolution
• 2 PCAP formats with 1 ns or 1000 ns resolution

TIME SYNCHRONIZATION
The Intel® PAC A10 GX time-stamping can be synchronized to OS time with dynamic drift adjustment.

FRAME PROCESSING
MULTI-PORT AND MULTI-CARD PACKET MERGE
It is possible to merge packets received on multiple 10G ports on the Intel® PAC A10 GX into a single data stream for further processing. This is useful for analyzing both receive and transmit directions of a connection, where the packets from the receive and transmit directions are received on separate ports. With Napatech Software Suite it is possible to merge data streams from multiple Intel® PAC A10 GX cards into a single data stream for analysis.

FRAME CLASSIFICATION
The Napatech Link Capture Software frame decoder ensures protocol recognition of all major layer 2 to 4 protocols. The frame classification information is provided together with each frame including offset information for encapsulated protocol header and payload data. This provides a significant processing offload and acceleration of analysis. Frame classification information can be used for:

• Finding protocol headers and payload data at dynamic locations
• Advanced filtering
• Dynamic selection of hash keys (per frame)
• Dynamic slicing
• Intelligent Multi-CPU distribution

FRAME AND PROTOCOL INFORMATION
For each Ethernet frame received, frame and protocol information is made available to the host application. This accelerates processing of each frame. Information available includes:

• High-precision time stamp information
• Protocol information: e.g. IPv4, IPv6, UDP, TCP, GRE, SCTP, EtherIP and GTP
• Encapsulation information: ISL, VLAN and MPLS
• Hash key information: hash key value and type
• Offsets to start of L3, L4 and L5 payload
• Coloring/tagging: Tags defined by filters
• Checksum error flags: Ethernet FCS

SLICING
With slicing it is possible to truncate frames so only essential information is analyzed. This reduces the processing load and thereby accelerates performance. Several types of slicing are supported:

• Fixed slicing: Truncation of frames to a maximum size or an offset from end of frame
• Dynamic slicing: Truncation of frames to a dynamic size relative to a specified protocol, e.g. IP payload + 16 bytes
• Conditional slicing: Fixed or dynamic slicing with properties based on frame decoding

ETHERNET FCS
For each received frame, the Ethernet Frame Check Sequence (FCS) is checked in hardware.

TUNNELING SUPPORT
GTP, IP-IN-IP, GRE, NVGRE AND VXLAN TUNNELING SUPPORT
Recognition of tunneling encapsulations enables efficient CPU load distribution based on the contents of the tunnel rather than the tunnel itself. This is especially useful when analyzing traffic on telecom backbone networks using GTP or IP-in-IP, or on enterprise networks using IP-in-IP, GRE, NVGRE or VxLAN.

INNER TUNNEL FRAME PROCESSING
For tunneled traffic it is important to perform analysis on the content of tunnels rather than the tunnel itself including methods for reducing the amount of traffic to be analyzed through slicing and filtering:

• Slicing of encapsulated packet headers and payloads
• Filtering on encapsulated packet headers and payloads

IP FRAGMENT HANDLING
Napatech Link Capture Software is capable of identifying fragmented IP packets and ensuring that associated fragments are distributed to the same host buffer for reassembly by the application. This helps accelerate the IP reassembly process for the application. Out-of-sequence fragments are also identified and handled by the Napatech Link Capture Software.

FLOW IDENTIFICATION
FLOW IDENTIFICATION BASED ON HASH KEYS
Frame classification information is used to calculate a hash value for each frame received. Frames with the same hash value can be treated as a flow and be processed accordingly.

MULTIPLE PROTOCOL-SPECIFIC HASH KEYS
Hash value calculation is based on combinations of specific header data. Napatech Link Capture Software supports up to 17 different hash key types based on the type of protocols encapsulated in the Ethernet frame. Protocol-specific hash keys allows flows to be established based on the type of protocol to be analyzed.

CUSTOM HASH KEYS
Custom hash keys can be configured of elements from frame headers and payload.
• 4 configurable byte-aligned offsets: 2 x 128 bits and 2 x 32 bits with separate bit masks
• IP protocol field from inner or outer IP header

DYNAMIC HASH KEY SELECTION
Since hash key types are protocol-specific, the Napatech Link Capture Software can dynamically select the correct hash key type for hash value calculation based on the information provided by the frame decoder.

CORRELATION KEY
With Napatech Link Capture Software it is possible to generate a correlation key that can be used to monitor individual packets at multiple points in the network. The correlation keyis a unique identifier for individual packets and can be used as an alternative to IP source and destination addresses for cases where network address translation can change IP addresses in the network being monitored. With correlation keys it is possible to measure the latency at multiple points in the network on a packet-by-packet basis. The correlation key can also be used for hardware acceleration of packet deduplication in application software. The 64-bit correlation key is generated in hardware and delivered to the application in the packet descriptor. The correlation key is calculated as a hash over configurable sections of the packet, and dynamic header information (e.g. TTL) can be masked out.

PACKET AND FLOW FILTERING
CONFIGURABLE GENERAL FILTERS
64 advanced programmable filters are available with an exceptional flexibility in the way they are specified, combined and controlled. They are configured by means of the easyto-use Napatech Programming Language (NTPL). The customer application can change the filters on the fly. The programmable filter logic is built on top of the advanced protocol decoding capabilities. This ensures that the application always gets the packets matching the requested protocol, even under very diverse conditions. Using a single filter, the Napatech Link Capture Software can capture, for instance, all TCP/IP packets with a specific source IP address, even when these packets are ISL-, VLAN- or MPLS-encapsulated and/or contain IP/TCP options.

PORT AND PROTOCOL FILTERS
Filtering can be performed based on the port on which frames are received and the protocol information from the frame decoder.

PATTERN COMPARE FILTERS
The pattern compare feature allows frames to be identified based on the application of a user-defined data pattern with bit masks at up to 4 specific offsets in the frame. A single frame can be compared against multiple data patterns.

IP GROUP MATCH FILTERS
Filtering can be performed based on match on groups of IP addresses. Up to 15 IP Group Match filters can be configured, sharing a pool of exact match entries and a pool of wildcard entries.
• Up to 36000 IPv4 or up to 5000 IPv6 exact address match entries
• Up to 864 IPv4 or up to 216 IPv6 wildcard address match entries

FLOW FILTERS
Filtering can be performed based on exact flow match.
• Up to 18000 IPv4 or up to 2500 IPv6 2-tuple flows
• Up to 12000 IPv4 or up to 2500 IPv6 5-tuple flows
• Up to 36000 IPv4 or up to 7500 IPv6 source or destination IP address match entries
• Configurable flow definitions, based on 2, 3, 4 or 5-tuple
• Flow definition selected per frame category (e.g. IP protocol or TCP port number)
• Up to 864 IPv4 or up to 216 IPv6 wildcard address match entries

FILTER ACTIONS
Filtering can result in the following actions:
• Packet discard, packet forward to a specific stream, packet coloring and packet slicing controlled per filter or flow by the application
• Flow ID delivered to application in packet descriptor

COLORING
Filtered frames can be tagged with a “color” ID identifying the filter that forwarded the frame. Color ID can be used to optimize applications performing different processing for different frame types.

INTELLIGENT MULTI-CPU DISTRIBUTION
CONFIGURABLE DATA DISTRIBUTION
Multi-CPU buffer splitting enables Napatech Link Capture Software to place captured frames in 1 to 64 host buffers. The customer can configure the size of the host buffers and how data is placed in the host buffers, based on results from the filter logic, port numbers and/or generated hash key values (flows).

The advanced multi-CPU buffer splitting functionality and the option for distributing traffic up to 64 CPU cores significantly improves the CPU cache performance, by always delivering the same flows and frame types to the same CPU cores.

HOST-BASED TRANSMIT
Multiple Tx streams enables multi-threaded processing and full throughput bidirectional Rx/Tx. The feature includes an API call for prepending protocol layers to Tx packets.

TRANSMIT ON TIME STAMP
Line rate transmit from application to network ports enables network simulation at 10G and 40G port speeds.
• Supports network test and simulation up to 40G line rate for any packet size
• Replay as captured based on time stamp
• Transmit frames immediately or according to nanoseconds time stamp
• Configurable rate limiter for control of network load
• Replay of PCAP files

LOCAL RETRANSMIT
Forwarding of traffic from one port to multiple ports eliminates the need for expensive load distribution switch ports and smart tap.
• Up to 40G line rate for any packet size
• Forwarding of traffic from any port to multiple ports within the same Intel® PAC A10 GX
• Fully flexible solution where packets received on one port input can be forwarded to any port output and where multiple port inputs can be forwarded in parallel
• Controls which packets are forwarded and on which ports based on packet classification and filter functions
• Supported for 10G and 40G speeds

COMBINED HOST TRANSMIT AND LOCAL RETRANSMIT
Enables session control and low latency for in-line use cases.
• Transmit from application to any port – simultaneous local retransmit between ports
• Typically, less than 5 microseconds latency for local retransmit between ports

Combined host transmit and local retransmit.

LOCAL RETRANSMIT WITH EMBEDDED TIME STAMP
Avoid use of expensive load distribution switch ports, while delivering high-precision time stamp to applications downlink.
• Preserve high-precision time stamp captured at tap point, when using Intel® PAC A10 GX for load distribution or replication to other appliances
• Trailer with time stamp appended to retransmitted frame
• Configurable enable/disable per output port

ADVANCED STATISTICS
RMON1 PORT STATISTICS
Napatech Link Capture Software provides RMON1 (RFC2819) counters on a per-port basis.

ADVANCED PORT STATISTICS
In addition to RMON1 port statistics, the Napatech Link Capture Software generates an extensive amount of additional statistics counters, which are available independently of whether the traffic is forwarded to the host or not. This enables customer applications to retrieve a comprehensive network traffic analysis with extremely low CPU load.

Two types of statistics counters are available:
• Large sets: RMON1 (RFC2819) counters with extension of jumbo frame counters are available for both captured and discarded frames on a per-port basis.
• Normal sets: Frame and byte counters for good and bad frames are available per color (filter) and per host buffer. Counter sets are always delivered as a consistent snapshot time-stamped by a 64-bit high-precision clock and can be synchronized with the Ethernet frames delivered for analysis.

MONITORING SENSORS
Sensors on the Intel® PAC A10 GX provide monitoring of FPGA temperature level with alarm and software shutdown.

Local retransmit with embedded time stamp support

Back To Top