Piggyback Node Metrics Dissemination in Server Clusters

Solution Overview

Problem

In distributed computing systems with heterogeneous nodes, existing methods for disseminating load data are inefficient, leading to bandwidth consumption and latency issues due to the need for frequent polling or periodic metric exchanges, which can overwhelm nodes and require additional security measures.

Innovation Solution

Implementing a 'piggyback' method where server nodes report metrics on transactional messages, allowing opportunistic dissemination of node metrics across the cluster without explicit polling, reducing inter-node communication and conserving bandwidth.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If frequent polling or periodic metric exchanges are used to disseminate load data, then nodes can maintain up-to-date information about cluster load, but bandwidth consumption increases and latency issues occur

Engineering Contradiction:

Improveload information freshnessVSAvoidbandwidth consumption

Core Design Contradiction:

Measurement precisionVSLoss of energy

Solution Approach 1:

The patent combines metric exchange communications with existing transactional messages between nodes. Instead of sending separate periodic polling requests, nodes piggyback load metric data on transactional messages that are already being exchanged for other purposes, thereby consolidating communication traffic and reducing overall bandwidth consumption while maintaining up-to-date load information.

Inventive Principle:

Principle #5Merging (Combining)

Solution Approach 2:

The patent introduces a metrics dissemination mechanism that acts as an intermediary by embedding load metrics within transactional messages. These messages serve as carriers that transport metric information between nodes without requiring dedicated communication channels, thus reducing the overhead associated with frequent polling while ensuring timely information exchange.

Inventive Principle:

Principle #24Intermediary (Mediator)

2Measurement precision

If frequent polling or periodic metric exchanges are used to disseminate load data, then nodes can maintain up-to-date information about cluster load, but latency increases

Engineering Contradiction:

Improveload information freshnessVSAvoidcommunication latency

Core Design Contradiction:

Measurement precisionVSLoss of time

Solution Approach 1:

The patent combines metric exchange communications with existing transactional messages between nodes. Instead of sending separate periodic polling requests, nodes piggyback load metric data on transactional messages that are already being exchanged for other purposes, thereby consolidating communication traffic and reducing overall bandwidth consumption while maintaining up-to-date load information.

Inventive Principle:

Principle #5Merging (Combining)

Solution Approach 2:

The patent leverages the continuous exchange of transactional messages between nodes to continuously disseminate load metrics. Since transactional messages are already being exchanged as part of normal cluster operations, the metric dissemination occurs continuously without requiring additional communication cycles, thus eliminating the latency associated with periodic polling while maintaining real-time load information freshness.

Inventive Principle:

Principle #20Continuity of useful action

3Measurement precision

If explicit polling methods are used for metric exchange, then load data can be collected systematically, but additional security measures are required and nodes can be overwhelmed

Engineering Contradiction:

Improveload data collectionVSAvoidsecurity measures and node overhead

Core Design Contradiction:

Measurement precisionVSDevice complexity

Solution Approach 1:

The patent combines metric exchange communications with existing transactional messages between nodes. Instead of sending separate periodic polling requests, nodes piggyback load metric data on transactional messages that are already being exchanged for other purposes, thereby consolidating communication traffic and reducing overall bandwidth consumption while maintaining up-to-date load information.

Inventive Principle:

Principle #5Merging (Combining)

Solution Approach 2:

The patent implements a self-service mechanism where nodes automatically include their load metrics in outgoing transactional messages and automatically process received metrics from other nodes. This eliminates the need for centralized polling coordination and reduces the overhead on individual nodes, as metric dissemination becomes an inherent part of normal transactional operations rather than an additional burden.

Inventive Principle:

Principle #25Self-service

Data Source

PatentUS10862957B2Dissemination of node metrics in server clusters

Publication Date: 2020.12.08 SALESFORCE INC

AI summary

In some embodiments, a server node may transmit metrics to a client piggy-backed on (appended to) responses to regular read/write requests. Other embodiments may be disclosed and/or claimed.