The SLink_31048_X is a high-performance L1+ ultra-low latency switch developed by Shengli Technologies specifically for the financial industry. Integrating the CrossPoint architecture with FPGA programmable technology, it has been developed for a variety of specialized applications targeting quantitative trading and high-concurrency market data processing scenarios.
The device utilizes an industry-leading FPGA chip solution and integrates multiple core functions: L1 data forwarding, picosecond-level high-precision timing, multi-group data aggregation, data filtering and isolation, and market data aggregation, fully meeting the financial sector's demand for extreme latency. Through hardware architecture innovation and software function customization, this product provides financial customers with a complete solution from nanosecond-level latency control to high-concurrency data processing. It is particularly suitable for time-sensitive core business scenarios such as low-latency trading and high-frequency market data analysis.
Built-in industry-leading FPGA chip
It has an industry-leading built-in Xilinx Virtex® UltraScale™ VU9P high performance FPGA chip
High bandwidth and large cache
Supporting up to 48 10G SFP+ ports, the selection of high-performance on-chip cache and high-bandwidth off-chip storage tailored to specific application characteristics. This meets the ultra-low latency and high throughput requirements of in high-frequency trading scenarios needed by brokerages, hedge funds, exchanges, and trading firms.
High precision clock
The switch's master clock can be selected from either an OCXO or Atomic source, ensuring a stable, high-precision clock even without an external clock source. Combined with Shengli Technologies’unique sampling and timing algorithms, it provides picosecond-level timestamp functionality, ensuring accurate capture time and the timeliness and accuracy of data analysis.
The SLink_31048_X low-latency switch supports three external clock synchronization methods, providing users with flexible time sync.
Support NTP clock synchronization
Support PTP clock synchronization
Support 1PPS input/output interface
Layer 1 Switch
SLink_31048_X supports flexible configuration of different application modules based on the diverse needs of financial scenarios.
As the underlying architectural cornerstone of the SLink_31048_X, the Layer1 Switch App uses CrossPoint matrix technology to build a 48-port pure Layer 1 switching framework, supporting 1G/10G dual-rate adaptive transmission. This module has the following technical features:
Ultra-low latency forwarding technology
/ Forwarding latency between any two ports is <4ns.
/ Message size has no impact on forwarding latency: Ethernet frames of any byte length are supported at line speed with zero jitter.
/ Utilizing industry-leading ultra-low latency components and technologies, ultra-low latency forwarding between ports is achieved.
Multifunctional data transmission and monitoring
/ Supports bidirectional patching for point-to-point links, enabling ultra-fast point-to-point data transmission.
/ Supports tapping for traffic replication using 1:N mapping links.
/ Supports signal regeneration and adaptive conversion between single-mode/multimode fiber and copper cable media.
/ Supports port traffic statistics and port status reporting.
Flexible configuration and management mechanism
/ Supports source MAC, allowing connections to upstream devices independent of downstream servers.
/ Supports per-port rate configuration.
/ Compatible architecture design enables Layer1 functionality to seamlessly coexist with other apps, such as network monitoring and traffic analysis.
NodeWatch precise traffic monitoring solution
As the core monitoring component of SLink_31048_X, NodeWatch App uses a hardware-level traffic capture engine to implement picosecond-level network behavior analysis, building a full-link visibility solution specifically for low-latency financial transaction scenarios.
“0” frame loss feature
/ NodeWatch utilizes a multi-dimensional architecture to achieve high-performance data transmission. Each network port is equipped with an independent high-concurrency cache unit, combining a 64GB large-capacity cache pool with 540Gb/s high-performance storage. Upload port rate limiting reduces the processing load on the receiving end, and jumbo frame slicing enables data compression and transmission for non-sensitive applications.
/ Dual-layer VLAN traversal is supported for cross-domain transmission. Full-port-level traffic statistics are provided, covering received/sent frame counts and error frame monitoring, creating a complete lossless transmission solution from cache management to traffic visualization.
High-precision hardware timestamp
/ 44 acquisition channels are completely homogenous, ensuring consistent timestamps
/ NodeWatch utilizes a multi-channel parallel architecture design, enabling full-channel parallel timestamping through dedicated hardware units, ensuring that every frame of data is instantly and accurately timestamped.
NodeMux multi-core aggregation engine technology analysis
As the low-latency data processing core of the SLink_31048_X, the NodeMux App achieves parallel aggregation and ultra-low latency forwarding of multi-channel data streams through a collaborative software and hardware architecture, specifically building high-concurrency data distribution capabilities for financial trading systems.
Multi-core independent processing architecture
/ NodeMux is a mode designed specifically for ultra-low latency aggregation. Downlink traffic is handled by L1 broadcast, while uplink aggregation is implemented by the FPGA.
/ Utilizing a switch matrix combined with FPGA programmable logic, multiple modes are selectable based on the number of cores (Mux Groups) and the number of ports supported by each core. Each Mux Group isolates end-to-end data flows, supporting dedicated aggregation paths for different market sources/trading channels.
/ External et ports are decoupled from internal ap ports, and the crosspoint matrix dynamically configures wiring methods for optimal cabinet deployment flexibility.
Extreme latency control technology
/ Uplink aggregate latency as low as 39ns
/ Downlink latency <4ns.
NodeAccess secure data isolation solution
As the core security component of SLink_31048_X in multi-tenant scenarios, the NodeAccess App integrates a hardware-level ACL filtering engine based on the NodeMux multi-core aggregation architecture, providing financial customers with nanosecond-level latency transmission and refined traffic control capabilities.
L1+ layered multiplexing technology
/ NodeAccess secure forwarding mode implements an N:1 multiplexing architecture based on the L1+ layer, and implements Mux and DeMux functions through FPGA hardware.
Refined ACL filtering engine
/ The SLink_31048_X switch integrates a multi-dimensional flow control engine, supporting flexible rule combinations based on MAC, IP, and protocol numbers. A single port can be configured with three sets of ALC rules for precise flow control.
/ The hardware-level filtering engine optimizes ACL filtering latency to 86ns. While ensuring data forwarding efficiency, the SLink_31048_X switch leverages collaborative filtering mechanisms at the MAC and protocol layers to achieve a balanced design between security policies and low latency. This design is particularly well-suited for scenarios such as financial trading, where flow control accuracy and latency are critical.
NodeAggregator high-concurrency market data aggregation solution
As the core processing module of the SLink_31048_X designed specifically for financial market scenarios, the NodeAggregator App uses a three-level cache architecture to achieve zero-frame-loss aggregation and nanosecond-level latency control for multi-channel, highly concurrent market data.
L1+Port-level exclusive cache
/ Each SFP+ port is equipped with 72KB of L1 cache, enabling data storage with extremely low latency.
Core-level shared cache
/ Each processing core is equipped with 1MB of L2 cache, supporting cross-port traffic load balancing and burst traffic absorption.
System-level massive cache
/ Each core integrates 16GB of L3 DDR4 cache, which can accommodate more than 10 million frames of market data.
NodeMuxSnap multi-functional integrated solution
As an innovative converged application of SLink_31048_X, the NodeMuxSnap App integrates the multi-core aggregation capabilities of NodeMux and the timestamp monitoring function of NodeWatch through software and hardware architecture integration, providing financial customers with the dual benefits of high-density port utilization and simplified deployment. Its technical features are presented in two core dimensions:
L1+Composite functional architecture design
/ Dual-mode working engine: Simultaneously supports Mux aggregation and Watch timestamping functions. Through FPGA logic partitioning technology, it achieves resource isolation and collaborative computing between the aggregation scheduling and timestamp processing modules. Port utilization optimization: A single device can complete the entire "aggregation-monitoring-forwarding" process, reducing the number of cable connections required for discrete devices in traditional solutions.
Balanced performance parameters
/ Aggregation performance: Combining the Mux and Watch functions, the data input by each core can be selectively timestamped while being aggregated from multiple channels. The Mux function and performance are not affected, and the Watch accuracy is 2ns.
