Phased Server Reservation Using Virtual Pools and Allotments
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing capacity management systems face inefficiencies due to low resource utilization and waste when reserving entire servers, exacerbated by supply-chain issues, and stackable server reservations introduce scalability challenges and break existing codebases.
Innovation Solution
A phased solving technique for stackable server reservations, comprising a server level allocation to estimate full server needs and an allotment level allocation to assign specific portions, using a virtual server pool and buffer servers to optimize resource allocation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If entire servers are reserved for capacity management, then service availability is ensured, but resource utilization decreases and server waste increases
Solution Approach 1:
The patent segments server resources into two levels: server-level allocation (reserving entire servers for reliability) and allotment-level allocation (dividing servers into smaller units called allotments for efficient resource distribution). This segmentation allows the system to maintain service availability through server reservations while improving resource utilization by allocating only the necessary portions of servers to specific workloads, thereby reducing waste of unused server capacity.
2Loss of energy
If stackable server reservations are implemented to improve resource utilization, then server waste is reduced, but system complexity increases and scalability is compromised
Solution Approach 1:
The patent divides the capacity management system into two distinct phases: server-level allocation (handling high-level server reservations) and allotment-level allocation (handling detailed resource distribution). This segmentation simplifies the system by separating concerns, making the codebase more maintainable and scalable while still enabling efficient resource utilization through allotments.
Solution Approach 2:
The patent introduces a new dimension of allocation granularity by creating allotments as intermediate units between entire servers and individual resources. This dimensional change allows the system to manage resources at multiple levels of abstraction, improving resource utilization without directly increasing complexity at any single level, as the allotment layer acts as a buffer that simplifies overall system management.
3Quantity of substance
If stackable server reservations are implemented to reduce server requirements, then resource efficiency improves, but existing codebases break and scalability is limited
Solution Approach 1:
The patent segments the allocation process into server-level and allotment-level phases, allowing existing codebases that work with entire servers to continue functioning at the server level, while new allotment-level functionality can be added without breaking existing systems. This phased approach maintains backward compatibility while enabling reduced server requirements through efficient allotment utilization.
Solution Approach 2:
The patent performs server-level allocation as a preliminary step before allotment-level allocation. This preliminary action establishes a foundation that existing codebases can work with, while subsequent allotment-level processing optimizes resource usage. By preparing the server-level structure first, the system enables reduced server requirements without disrupting existing codebases that operate at the higher server level.
Data Source
AI summary
A method of resource allocation is disclosed. A request from a user to allocate computing resources for specified compute processing is received. A server level allocation is performed, including by determining a full server reservation estimate corresponding to the request, wherein the full server reservation estimate comprises a count of how many servers are reserved for the specified compute processing. A virtual server pool is formed based at least in part on the full server reservation estimate. An allotment level allocation is performed, including by assigning among the virtual server pool one or more allotments for the specified compute processing.