Deep packet inspection (DPI) is a powerful technology that analyzes packets of data traveling across the internet. Many law enforcement agencies use DPI systems to implement lawful surveillance and intercept programs to gather evidence in criminal investigations and help prevent terrorism.
Today’s electronically-connected world leaves societies and businesses open to network attacks. Being able to quickly capture and analyze packets without disrupting other server operations is possible with an HPN. With a programmable NIC, analytics can be performed on packets at the NIC, greatly reducing CPU overhead while delivering extreme performance. At the same time, extreme performance for packet injection allows entities to perform testing to ensure their infrastructure can withstand any attacks, such as distributed denial of service (DDoS).
For these markets, Emulex provides Network Xceleration (NX) solutions which require the Emulex OneConnect® OCe12000-D family of 10GbE network adapters and FastStack™ software. Deep packet inspection in an HPN is achieved with FastStack Sniffer10G™.
FastStack Sniffer10G delivers High Performance Networking with extreme high performance network packet capture, injection and analysis, all using the standard 10GbE network. With industry-leading performance, FastStack Sniffer10G provides higher throughput, low latency, enhanced multi-core awareness, precision timestamping and port-aggregation and merging. Figure 3 shows how FastStack Sniffer10G fits into an HPN.
High Performance Networking drives accelerated network performance with FastStack Sniffer10G kernel bypass operations.
FastStack Sniffer10G delivers predictable and reliable performance to ensure no packet loss impacts business operations. Supporting the maximum achievable packet rate of 14.88 million packets per second, this solution is specifically designed for applications where high throughput is required.
In summary, FastStack Sniffer10G with OneConnect OCe12000-D adapters provides:
- High throughput: Full 100 percent packet capture for all Ethernet packet sizes (not optimized for 64-byte or other certain sweet spots)
- Userspace (OS-bypass) data paths: Reduce copying and synchronization costs incurred by traditional kernel-based receive operations and optimized for user applications tightly coupled with data
- Multi-core awareness: Can use all CPU cores for packet analysis with:
- Partitioning with flow awareness: logical TCP/UDP flows can be received into multiple receive rings (or buffers)
- Duplication that allows multiple applications to view the traffic
- Provides features typically available only in highly specialized and high cost Ethernet appliances
- Implementer's Lab blog:
Interop 2012 - Emulex New Product Demos at the Emulex Booth
- Emulex Labs blog:
Introducing New I/O Solutions for the High Performance Computing Market
HFT-derived volumes now account for half or more of trading on exchanges such as the London Stock Exchange and U.S. trading platforms (Reuters article, “Ultra fast trading needs curbs -global regulators,” July 7, 2011)