Fault Tolerant Architecture for Low-Latency Transaction Processing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing electronic trading systems face challenges in achieving fault tolerance with minimal recovery time and performance degradation due to network interconnect latencies and the need for geographically disparate backup systems, which introduce communication asymmetries and increased latency.
Innovation Solution
An integrated component system is implemented on a single processing platform using shared memory communication and inter-process communication to eliminate network interconnects, with a fault tolerance mechanism that synchronizes and designates backup instances to ensure seamless failover without network latency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If backup systems are located in geographically disparate regions to ensure fault tolerance, then system reliability is improved, but communication latency increases
Solution Approach 1:
The patent transitions from geographic distribution (spatial dimension) to logical distribution within a single data center (temporal/organizational dimension). By placing primary and backup systems in the same physical location but isolating them through virtualization and logical networking, the system achieves fault tolerance without the communication latency penalties of geographic distribution.
Solution Approach 2:
The patent introduces intermediary components including virtual network interfaces, message queues, and state synchronization mechanisms that mediate between primary and backup systems. These intermediaries enable reliable state replication and failover while maintaining low latency through optimized communication paths within the same data center.
2Adaptability or versatility
If network interconnects are used to connect system components, then system modularity and deployment flexibility are improved, but communication latency increases
Solution Approach 1:
The patent merges network communication functions with local inter-process communication by implementing message queues and state synchronization mechanisms that operate within the same host or data center. This combines the benefits of network modularity with the speed of local communication, eliminating unnecessary network hops while maintaining system flexibility.
Solution Approach 2:
The patent creates copied instances of system components (primary and backup trading systems) that run in isolated virtual environments. These copies communicate through optimized channels including shared memory and message queues, replicating the functionality of networked systems while achieving lower latency through proximity.
3Loss of time
If fully redundant backup components are implemented, then recovery speed is improved, but system cost increases
Solution Approach 1:
The patent implements partial redundancy where backup systems maintain synchronized state data but do not continuously process all inputs. The backup receives and processes a subset of inputs or processes them asynchronously, achieving sufficient redundancy for fast failover without the full resource overhead of complete active duplication.
Solution Approach 2:
The patent performs preliminary state synchronization continuously in the background before failures occur. By maintaining pre-synchronized state data through message queues and state replication mechanisms, the system prepares for rapid failover without requiring full real-time processing redundancy, reducing ongoing resource requirements.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A system/process provides fault tolerance to an integrated component system which integrates core components of a transaction processing system into a single processing platform, i.e., a single server, enabling elimination of the network interconnects and associated latencies introduced thereby in favor of much faster interconnects, such as interprocess communication and shared memory communication messaging, where a failure of any one component necessitates failing over the entire system to a backup thereof.