Distributed DNS Reflection for Accurate RTT Estimation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for estimating round-trip times (RTTs) between Local Domain Name Servers (LDNSs) and data centers are often inaccurate due to variations in network traffic and resource availability, limiting the effectiveness of Global Load Balancing (GLB) in selecting optimal data centers for client requests.
Innovation Solution
The implementation of a method using distributed DNS reflection, where data centers share a shared database, including a distributed ledger, to proactively estimate and improve RTT values by sending requests between reflectors and collectors, and calculating improved RTT values using timestamps and latency measurements.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If active methods (ping or query) are used to estimate RTTs between LDNSs and data centers, then RTT estimates can be obtained, but reachability is limited as many data centers may not respond to such pings or queries due to firewalls that drop packets
Solution Approach 1:
The patent uses DNS queries as an intermediary mechanism to estimate RTTs. Instead of directly pinging data centers (which may be blocked by firewalls), the system sends DNS queries through the normal DNS resolution path, allowing RTT measurement without triggering firewall drop rules. The DNS infrastructure acts as a mediator that enables measurement while bypassing security restrictions.
2Productivity
If rough relative estimates of RTTs are obtained based on geographic proximity information, then data centers physically located closer to a particular LDNS are assumed to have relatively shorter RTTs, but variations in network traffic, changes in network topology, and variations in computing resource loads make such estimates inaccurate
Solution Approach 1:
The system performs preliminary RTT measurements by sending DNS queries to data centers before actual client requests arrive. These advance measurements populate a cache of RTT values that can be quickly retrieved during GLB decisions, avoiding the need for real-time measurements while maintaining accuracy. The measurements are taken in advance when network conditions are more predictable.
Solution Approach 2:
The system implements feedback mechanisms where RTT measurements are continuously updated and stored. When a data center is selected to service requests, its RTT measurement is fed back into the system and used to update future routing decisions. This creates a closed-loop system that adapts to changing network conditions and resource loads over time, improving accuracy while maintaining fast decision-making.
Data Source
AI summary
Round-trip times (RTT) values are estimated between meshed data centers serving an internet domain and a local domain name server (LDNS) associated with the data centers. A method initializes a shared database with proactively estimated RTTs and uses distributed DNS reflection to improve the accuracy of an estimated RTT value and update the shared database. RTTs may be proactively estimated by using pings, queries, or local DNS reflection. In local DNS reflection, a reflector sends a request to a collector in the same data center via the LDNS, and the collector measures the RTT. In distributed DNS reflection, a reflector in a first data center sends a request to a collector in a second data center via the LDNS. The collector calculates the latency, retrieves the (first) RTT between the first data center and the LDNS, and calculates the (second) RTT between the second data center and the LDNS by doubling the difference between the latency and half the first estimated RTT.