Workload Distribution in Virtualized Systems

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

In scalable computing environments, the variability in booting and initializing virtual machines leads to instability and revenue loss due to inefficient workload management, particularly in handling heterogeneous workloads and services with stringent Service Level Agreements (SLAs).

Innovation Solution

A method and system that orchestrate resources by decoupling services from virtual instances, combining multiple workloads to reduce reconfigurations through normalization, smoothing, workload categorization, and intelligent distribution, using a Reconfiguration Manager to instantiate or delete VMs, thereby minimizing reconfiguration costs and improving server utilization.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Adaptability or versatility

If virtual machines are instantiated or deleted to accommodate workloads, then workload management flexibility is improved, but system stability deteriorates due to variability in booting and initializing times

Engineering Contradiction:
Improveworkload management flexibilityVSAvoidsystem stability
Core Design Contradiction:
Adaptability or versatilityVSReliability

Solution Approach 1:

The patent applies preliminary action by pre-booting virtual machines before workloads are assigned to them. The system identifies workloads that need to be moved and proactively boots target virtual machines in advance, so they are ready when the workload migration is needed. This eliminates the variability and delay associated with booting virtual machines at the moment of workload assignment, thereby maintaining system stability while preserving workload management flexibility.

Inventive Principle:
Principle #10Preliminary action

2Ease of operation

If multiple workloads are handled individually, then workload-specific management is improved, but reconfiguration costs increase due to frequent VM instantiations

Engineering Contradiction:
Improveworkload-specific managementVSAvoidreconfiguration time
Core Design Contradiction:
Ease of operationVSLoss of time

Solution Approach 1:

The patent applies merging by combining multiple workloads and assigning them to the same virtual machine when possible. The system evaluates whether workloads can be co-located on shared virtual machines, reducing the total number of virtual machine instantiations and deletions required. This approach maintains individual workload management capabilities while significantly reducing reconfiguration time and operational overhead by handling multiple workloads collectively rather than individually.

Inventive Principle:
Principle #5Merging (Combining)

3Reliability

If virtual machines are frequently instantiated to meet demand, then service availability is improved, but resource utilization efficiency deteriorates

Engineering Contradiction:
Improveservice availabilityVSAvoidresource utilization efficiency
Core Design Contradiction:
ReliabilityVSProductivity

Solution Approach 1:

The patent applies universality by creating virtual machines that can serve multiple workloads and services simultaneously. Instead of instantiating dedicated virtual machines for each workload, the system creates multi-functional virtual machines that can be dynamically assigned to different workloads based on demand. This approach maintains high service availability by ensuring virtual machines are ready when needed, while improving resource utilization efficiency by avoiding the overhead of frequently creating and destroying single-purpose virtual machines.

Inventive Principle:
Principle #6Universality (Multi-functionality)

Data Source

PatentEP3291091B1A method and system for managing workload distribution in a virtualized system
Publication Date: 2022.11.30 ALCATEL LUCENT SA
  • EP3291091B1 patent drawingFigure 1
  • EP3291091B1 patent drawingFigure 2
  • EP3291091B1 patent drawingFigure 3

AI summary

A method for managing workload distribution in a system, the method comprising for a set of services executed in the system, categorising workloads to generate at least one workload group comprising a set of workloads with a combined reconfiguration cost less than the sum of individual reconfiguration costs of the selected workloads and distributing workloads of the generated workload group to the set of services according to the type of respective ones of the set of services.