Resource Placement Manager Jam Placement Strategy
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In virtualized computing environments, the placement of storage and compute resources relative to each other can impact performance, and existing systems struggle to ensure performance guarantees due to constraints like infrastructure diversity, durability, and capacity limitations, leading to situations where resource instances cannot be placed in optimal domains due to capacity constraints.
Innovation Solution
Implementing a resource placement manager that uses APIs to place resource instances in the same placement domain as the instance they are attached to, while allowing for 'jam' placements that temporarily violate constraints, followed by mitigating actions to restore compliance with constraints, such as migrating non-critical instances to other domains.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If resource instances are placed in the same placement domain to meet performance guarantees, then performance is improved, but capacity constraints are violated
Solution Approach 1:
The system dynamically adjusts placement constraints based on current capacity conditions. When capacity allows, strict placement constraints are enforced to ensure performance guarantees. When capacity is exceeded, the system temporarily relaxes constraints through 'jam placement' and subsequently restores compliance by migrating non-critical instances, making the constraint enforcement dynamic rather than static.
Solution Approach 2:
The system changes the state of placement constraints from enforced to relaxed temporarily during jam placement scenarios. By modifying the constraint parameter state based on capacity conditions, the system can place resource instances in optimal domains when needed while allowing temporary violations when capacity is exceeded, then restoring the original constraint state after mitigation.
2Quantity of substance
If placement constraints are strictly enforced, then capacity constraints are satisfied, but resource instances cannot be placed in optimal domains
Solution Approach 1:
The system performs preliminary assessment of capacity conditions before placing resource instances. When capacity constraints would prevent optimal placement, the system proactively identifies this condition and prepares to apply jam placement with subsequent mitigation actions, rather than simply failing the placement request.
Solution Approach 2:
The system converts the harmful effect of capacity constraints (which would normally prevent optimal placement) into a benefit by implementing a controlled violation mechanism. The temporary constraint violation enables optimal placement, and the subsequent automated mitigation restores capacity balance, ultimately improving overall system performance while maintaining long-term capacity compliance.
3Reliability
If jam placement is performed to place resource instances in optimal domains, then performance guarantees are met, but placement constraints are temporarily violated
Solution Approach 1:
The system implements continuous feedback monitoring of placement constraint compliance. After performing jam placement that temporarily violates constraints, the system detects the violation state and triggers automated mitigation actions to restore compliance. This feedback loop ensures that constraint violations are temporary and self-correcting, maintaining operational simplicity while achieving performance guarantees.
Data Source
AI summary
A distributed system may include one or more interfaces, such as APIs, allowing a client to request a resource instance and further provide information about the resource instance that can be used to determine a placement domain for resource instance. For example, a compute instance may be placed in a same placement domain as a volume that provides block storage for the compute instance. In some embodiments, a resource placement manager may “jam” place a resource instance in a particular placement domain, wherein the “jam” placement causes one or more placement constraints to be temporarily violated. The resource placement manager may then perform one or more mitigating actions to return the particular placement domain to a state that satisfies the one or more placement constraints.