Cloud Storage Tiering via SDN Abstraction for Latency Control
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing storage tiering methods do not account for network latency, which can lead to performance issues and violations of service level agreements in cloud or data center environments, especially where compute nodes and storage nodes are allocated to different tenants on demand and connected via converged networks.
Innovation Solution
The solution involves using Software Defined Network (SDN) technology, specifically OpenFlow, to abstract storage nodes into tiered resource pools with virtual IP addresses, automatically provisioning high performance storage close to application nodes and migrating storage as needed to maintain low latency, even when applications are moved to new locations, utilizing a center control plane and agents to manage storage and network devices.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If storage nodes are allocated to different tenants on demand via converged networks, then resource utilization and adaptability are improved, but network latency increases causing performance degradation
Solution Approach 1:
The system segments storage resources into performance-based storage nodes and capacity-based storage nodes organized in tiered resource pools. This segmentation allows compute nodes to access appropriate storage tiers based on performance requirements while maintaining flexible tenant allocation, thereby reducing network latency for performance-critical operations.
Solution Approach 2:
The patent introduces an intermediary layer (storage resource pool abstraction with virtual addresses) between compute nodes and physical storage nodes. This intermediary enables flexible tenant allocation while managing network latency by routing requests through optimized paths and maintaining virtual address mappings that reduce direct network dependency.
2Loss of energy
If storage tiering is implemented without considering network latency, then storage cost is reduced, but service level agreement compliance deteriorates
Solution Approach 1:
The system implements dynamic storage tiering that adapts to network conditions and workload requirements. Storage nodes can be dynamically allocated between performance-based and capacity-based tiers based on real-time demands, ensuring service level agreement compliance while optimizing storage cost through automated tier migration.
Solution Approach 2:
The patent changes the parameter of storage allocation by introducing performance-based versus capacity-based storage node classification. This parameter change enables the system to balance cost and performance by allocating appropriate storage types to different tenants and workloads based on service level requirements.
3Speed
If compute nodes access storage nodes directly without abstraction, then access speed is improved, but system complexity and adaptability to tenant movements worsen
Solution Approach 1:
The patent creates a universal storage resource pool abstraction that serves multiple functions: it provides fast access through performance-based nodes, enables flexible tenant allocation, simplifies management through centralized control, and maintains adaptability to tenant movements. This multi-functional abstraction resolves the contradiction between access speed and system complexity.
Data Source
AI summary
A plurality of performance-based storage nodes and a plurality of capacity-based storage nodes of a data storage system in a network environment are allocated to one or more tiered resource pools such that the performance-based storage nodes and the capacity-based storage nodes allocated to each one of the one or more tiered resource pools are addressable via a given virtual address for each tiered resource pool. Access to the performance-based storage nodes and the capacity-based storage nodes in the one or more tiered resource pools by a plurality of compute nodes is managed transparent to the compute nodes via a given storage policy. At least portions of the compute nodes, the performance-based storage nodes, and the capacity-based storage nodes are operatively coupled via a plurality of network devices. One or more of the allocating and managing steps are automatically performed under control of at least one processing device.