Memory Sled Filtering for Data Center Query Efficiency

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Conventional data centers face inefficiencies and performance drag due to the need to move large volumes of data for composite queries, as they often transfer entire data sets across multiple nodes, which is inefficient and resource-intensive.

Innovation Solution

Implementing memory sleds within data centers that can filter data sets and provide filtered results to compute sleds, utilizing memory filtering logic to execute queries and reduce data transfer by only transmitting the necessary data.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If entire data sets are transferred across multiple nodes for composite queries, then query results can be obtained, but data transmission volume increases and system efficiency decreases

Engineering Contradiction:
Improvequery processing efficiencyVSAvoiddata transmission volume
Core Design Contradiction:
ProductivityVSQuantity of substance

Solution Approach 1:

The patent applies preliminary action by executing filter operations at the memory sled before data transfer. The memory filtering logic evaluates predicates and filters the data set in advance, so that only the necessary filtered results are transmitted to compute sleds. This prevents the need to transfer entire data sets, thereby reducing data transmission volume while maintaining query processing efficiency.

Inventive Principle:
Principle #10Preliminary action

2Quantity of substance

If filter operations are executed at memory sleds, then data transmission is reduced, but additional memory filtering logic is required

Engineering Contradiction:
Improvedata transmission volumeVSAvoidmemory filtering logic
Core Design Contradiction:
Quantity of substanceVSDevice complexity

Solution Approach 1:

The patent applies universality by designing memory sleds with multi-functionality. The memory sleds not only store data but also execute filter operations through integrated memory filtering logic. This allows the same hardware component to perform both storage and filtering functions, reducing the need for separate filtering infrastructure and minimizing the increase in device complexity.

Inventive Principle:
Principle #6Universality (Multi-functionality)

Solution Approach 2:

The patent applies self-service by enabling memory sleds to autonomously perform filter operations on their stored data sets. The memory filtering logic resides within the memory sled itself, allowing it to independently evaluate predicates and filter data without requiring external processing. This self-service capability reduces data transmission needs while keeping the added complexity localized within existing memory sled components.

Inventive Principle:
Principle #25Self-service

Data Source

PatentEP3474149B1Memory filtering for disaggregate memory architectures
Publication Date: 2021.09.22 INTEL CORP
  • EP3474149B1 patent drawingFigure 1
  • EP3474149B1 patent drawingFigure 2
  • EP3474149B1 patent drawingFigure 3

AI summary

Examples may include a data center in which memory sleds are provided with logic to filter data stored on the memory sled responsive to filtering requests from a compute sled. Memory sleds may include memory filtering logic arranged to receive filtering requests, filter data stored on the memory sled, and provide filtering results to the requesting entity. Additionally, a data center is provided in which fabric interconnect protocols in which sleds in the data center communicate is provided with filtering instructions such that compute sleds can request filtering on memory sleds.