Network Switch Attached Storage for Bandwidth Reduction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing computing systems face high network bandwidth consumption due to frequent data transfer between host servers and remote data storage devices, especially when datasets are spread across multiple local storage devices, leading to reduced effective bandwidth and increased network traffic.
Innovation Solution
Storing datasets on data storage devices directly attached to network switches, which are strategically chosen to minimize network hops and reduce bandwidth usage, allowing workloads to access data with fewer network requests, and using local storage when capacity is sufficient.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If datasets are stored on remote data storage devices accessed over the network, then storage capacity is increased, but network bandwidth consumption increases and effective bandwidth decreases
Solution Approach 1:
The patent implements software-defined storage that selectively places datasets in different locations based on local storage availability and workload requirements. When local storage on host servers is sufficient, datasets are stored locally to eliminate network bandwidth consumption. When local storage is insufficient, datasets are stored on network-attached storage devices. This local-quality approach ensures that storage capacity is maximized while network bandwidth is consumed only when absolutely necessary.
Solution Approach 2:
The patent introduces network-attached storage devices as intermediary storage resources between host servers with insufficient local storage and the network. These intermediary storage devices are directly attached to network switches, providing a middle ground that increases overall storage capacity while minimizing network bandwidth consumption by being strategically positioned closer to multiple host servers simultaneously.
2Loss of energy
If datasets are stored on local storage devices of host servers, then network bandwidth consumption is reduced, but storage capacity is limited by individual server capacity
Solution Approach 1:
The patent creates a universal storage system where network-attached storage devices serve multiple host servers simultaneously. A single network-attached storage device can be shared by multiple workloads on different host servers, providing universal access to the same storage resource. This multi-functionality allows the system to pool storage capacity across multiple servers, effectively increasing total storage capacity while maintaining low network bandwidth consumption through strategic placement.
3Productivity
If datasets are migrated between physical hosts, then workload balancing is improved, but data access performance deteriorates due to network transfers
Solution Approach 1:
The patent implements software-defined storage that maintains local copies of datasets on host servers when possible, allowing workloads to be migrated between physical hosts without requiring physical data migration. The system dynamically determines the optimal location for each dataset based on current workload placement, ensuring that data access performance is maintained even as workloads move between hosts for load balancing purposes.
4Reliability
If dedicated network-attached storage devices are used, then data access uniformity is improved, but system complexity and cost increase
Solution Approach 1:
The patent repurposes existing network switches to provide storage functionality by directly attaching storage devices to them. This transforms the network switch from a simple networking device into a multi-functional device that also provides storage services. This approach eliminates the need for dedicated network-attached storage devices, reducing system complexity and cost while maintaining uniform data access performance through the network switch's central position in the network architecture.
Data Source
AI summary
A computer program product provides a non-volatile computer readable medium and non-transitory program instructions embodied therein, where the program instructions are configured to be executable by a processor to cause the processor to perform various operations. The operations may comprise identifying a dataset that is being used by a first workload running on a first host server within a computing system and a second workload running on a second host server within the computing system, wherein each of the host servers within the computing system are connected for communication over a network including a network switch. The operations may further comprise identifying a data storage device that is directly attached to the network switch within the network, migrating the identified dataset to the identified data storage device, and causing the first and second workloads to send dataset access requests to the identified dataset on the identified data storage device.