Burst Buffer Middleware Appliance for Batch IO Aggregation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In parallel computing systems, the high volume of metadata generates numerous small input-output (IO) requests, leading to increased costs and inefficiencies in multi-tier storage environments, particularly as storage tiers become more distant, necessitating a method to aggregate multiple IO operations into batch network operations.
Innovation Solution
A multi-tier storage system employing a burst buffer middleware appliance with an abstract storage interface that provides a list API, enabling batch operations to be processed simultaneously, thereby leveraging emerging storage devices' list APIs and reducing the cost of individual IO operations, and supporting legacy storage devices without list APIs by converting operations into individual requests.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If multiple small IO operations are used for metadata in multi-tier storage environments, then data can be stored and accessed, but the cost and inefficiency increase significantly as storage tiers become more distant
Solution Approach 1:
The patent merges multiple individual IO operations into a single batch operation. The burst buffer appliance collects multiple metadata read/write requests and consolidates them into one batch request that is sent to the persistent storage device in a single network round trip, thereby reducing the cumulative round trip cost and improving efficiency.
Solution Approach 2:
The burst buffer appliance acts as an intermediary between the parallel computing system and the persistent storage device. It receives individual IO operations from the compute nodes, aggregates them into batches, and sends consolidated requests to the storage device, thereby mediating the communication and reducing the number of round trips required.
2Productivity
If batch operations are implemented using a burst buffer appliance, then multiple IO operations can be processed simultaneously reducing round trip cost, but device complexity increases
Solution Approach 1:
The burst buffer appliance serves as a dedicated intermediary component that handles the complexity of batch aggregation. By isolating the batch processing logic in this separate appliance, the overall system architecture remains modular and manageable, with the burst buffer absorbing the complexity of coordinating multiple IO operations.
Solution Approach 2:
The storage system is segmented into distinct functional components: compute nodes, burst buffer appliances, and persistent storage devices. This segmentation allows the burst buffer to specialize in batch aggregation while storage devices focus on data persistence, distributing complexity across dedicated components rather than concentrating it everywhere.
3Ease of manufacture
If list API is used for batch operations, then storage device cost is reduced, but compatibility with legacy storage devices without list API becomes problematic
Solution Approach 1:
The burst buffer appliance acts as a compatibility intermediary between applications using list API and legacy storage devices that do not support it. It translates list API batch operations into individual IO operations that legacy devices can understand, while still providing the performance benefits of batching for compatible storage systems.
Solution Approach 2:
The system dynamically adjusts the operation mode based on the storage device capabilities. When a legacy device without list API is detected, the burst buffer changes its behavior to use individual operations instead of batched list operations, allowing the same architecture to adapt to different storage device parameter sets.
Data Source
AI summary
A multi-tier storage system is provided that employs a burst buffer middleware appliance for batch messaging. An exemplary multi-tier storage system comprises at least first and second storage tiers comprising storage for storing data; and at least one burst buffer appliance, wherein the at least one burst buffer appliance comprises an abstract storage interface comprising a software interface module that communicates with at least one persistent storage device in at least one of the first and second storage tiers, wherein the at least one persistent storage device provides a list application programming interface, wherein the software interface module provides a batch operation comprised of a plurality of individual operations to the at least one persistent storage device using an aggregated message. The batch operation optionally comprises a batch operation for processing metadata. A burst buffer appliance is also provided for implementing the batch messaging techniques described herein.


