Geo-site-aware Cluster Workload Placement

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current cluster systems are not geo-site-aware, leading to inefficient workload distribution and failover decisions, as they do not natively recognize the geographical location of cluster nodes, resulting in suboptimal resource usage and increased network traffic costs.

Innovation Solution

Implementing a geo-site-aware cluster system that uses geo-site awareness to manage node placement, failover, and quorum management, by identifying physical and virtual resources at each geo-site, optimizing workload placement, and making decisions based on geographical locations to reduce inter-site traffic and enhance resilience.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If cluster systems distribute workloads based on conventional factors like CPU or memory utilization, then resource utilization is optimized, but geo-site efficiency deteriorates resulting in increased network traffic costs and latency

Engineering Contradiction:
Improveresource utilizationVSAvoidnetwork traffic cost
Core Design Contradiction:
ProductivityVSLoss of energy

Solution Approach 1:

The patent applies local quality by making the cluster system geo-site-aware, where workload placement decisions are made locally based on the geo-site of the client rather than uniformly across all nodes. The system identifies which subset of cluster nodes is located at which geo-site and places workloads locally at the client's geo-site, reducing the need for cross-geo-site network traffic while maintaining efficient resource utilization.

Inventive Principle:
Principle #3Local quality

2Adaptability or versatility

If cluster systems manually specify node subsets for geo-locations, then geo-site awareness is partially achieved, but system complexity and administrative overhead increase

Engineering Contradiction:
Improvegeo-site awarenessVSAvoidconfiguration complexity
Core Design Contradiction:
Adaptability or versatilityVSDevice complexity

Solution Approach 1:

The patent implements self-service by enabling the cluster system to automatically discover and identify geo-sites and the subset of cluster nodes located at each geo-site without manual configuration. The system natively recognizes geo-locations and uses this information intelligently in cluster configuration, workload placement, and quorum maintenance, eliminating the need for administrators to manually specify node subsets.

Inventive Principle:
Principle #25Self-service

3Productivity

If cluster systems place workloads and storage at different geo-sites, then resource distribution is maximized, but system resiliency and performance deteriorate due to increased network dependency

Engineering Contradiction:
Improveresource distributionVSAvoidsystem resiliency
Core Design Contradiction:
ProductivityVSReliability

Solution Approach 1:

The patent applies local quality by ensuring that both workloads and their dependent storage are placed at the same geo-site. This local co-location strategy reduces dependency on cross-geo-site network links, improving system resiliency by minimizing the impact of network failures while maintaining efficient resource distribution across multiple geo-sites.

Inventive Principle:
Principle #3Local quality

4Reliability

If cluster systems use conventional failover mechanisms without geo-site awareness, then failover capability is provided, but failover decisions are suboptimal leading to increased recovery time

Engineering Contradiction:
Improvefailover capabilityVSAvoidrecovery time
Core Design Contradiction:
ReliabilityVSLoss of time

Solution Approach 1:

The patent implements feedback by using geo-site awareness to inform failover decisions. The system monitors the geo-site locations of cluster nodes and uses this information to make intelligent failover decisions, preferring to failover within the same geo-site to minimize recovery time and maintain performance, rather than relying on conventional failover mechanisms that lack geo-site context.

Inventive Principle:
Principle #23Feedback

Data Source

PatentEP3338186B1Optimal storage and workload placement, and high resiliency, in geo-distributed cluster systems
Publication Date: 2023.03.29 MICROSOFT TECHNOLOGY LICENSING LLC
  • EP3338186B1 patent drawingFigure 1
  • EP3338186B1 patent drawingFigure 2
  • EP3338186B1 patent drawingFigure 3

AI summary

Technologies for cluster systems that are natively geo-site-aware. Such a cluster system makes use of this awareness to determine the subsets of nodes located at various geo-sites at physical configuration, to optimize workload placement based on the geo-sites, to make failover and failback decisions based on the geo-sites, and to assign voting and prune nodes for quorum management based on the geo-sites. Such capabilities result in cluster systems that are more resilient and more efficient in terms of resource usage than cluster systems without such native geo-site awareness.