Cryptographic Metadata Compression for Low-Latency Memory Encryption

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current memory encryption engines face significant bandwidth and latency issues due to the frequent transfer of cryptographic metadata between on-die and off-die memory during enclave operations, particularly in multi-threaded server architectures, which also leads to increased power consumption.

Innovation Solution

Implementing a compression-decompression engine within the memory encryption engine to compress cryptographic metadata, such as counter values and MACs, and storing them in a dedicated cache to reduce the need for off-die memory access, using techniques like run-length encoding and dictionary schemes to optimize storage and retrieval.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If cryptographic metadata is frequently transferred between on-die and off-die memory during enclave operations, then memory encryption security is maintained, but bandwidth requirements and latency increase significantly

Engineering Contradiction:
Improvememory encryption securityVSAvoidlatency
Core Design Contradiction:
ReliabilityVSLoss of time

Solution Approach 1:

The patent pre-loads cryptographic metadata (counter values and MACs) into an on-die buffer before they are needed for encryption operations. This preliminary action allows the memory encryption engine to access metadata from the fast on-die buffer rather than repeatedly accessing off-die memory, thereby reducing latency while maintaining security requirements.

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The patent introduces an intermediary buffer structure located on-die that sits between the off-die memory and the memory encryption engine. This buffer acts as a mediator, caching frequently accessed cryptographic metadata and reducing the need for direct off-die memory accesses, thus lowering latency without compromising the integrity of the encryption process.

Inventive Principle:
Principle #24Intermediary (Mediator)

2Reliability

If cryptographic metadata is frequently transferred between on-die and off-die memory, then memory encryption functionality is maintained, but bandwidth requirements increase

Engineering Contradiction:
Improvememory encryption functionalityVSAvoidbandwidth
Core Design Contradiction:
ReliabilityVSQuantity of substance

Solution Approach 1:

The system pre-loads cryptographic metadata into the on-die buffer in advance, so that subsequent encryption operations can proceed using the cached metadata without requiring additional bandwidth for real-time off-die memory accesses. This reduces the overall bandwidth consumption while maintaining full encryption functionality.

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The on-die buffer serves as an intermediary that captures and retains cryptographic metadata, preventing repeated traversals of the off-die memory interface. This intermediary structure significantly reduces bandwidth requirements by localizing metadata access to the fast on-die memory subsystem.

Inventive Principle:
Principle #24Intermediary (Mediator)

3Quantity of substance

If cryptographic metadata is stored in off-die memory, then sufficient storage capacity is available, but power consumption increases due to frequent transfers

Engineering Contradiction:
Improvestorage capacityVSAvoidpower consumption
Core Design Contradiction:
Quantity of substanceVSUse of energy by moving object

Solution Approach 1:

The system proactively loads cryptographic metadata into the on-die buffer before encryption operations commence, eliminating the need for repeated power-consuming transfers between off-die memory and the encryption engine. This preliminary loading action reduces overall power consumption while maintaining adequate storage capacity through the buffer.

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The on-die buffer acts as a power-saving intermediary by intercepting metadata access requests and serving them from local storage rather than triggering high-power off-die memory transfers. This intermediary approach significantly reduces power consumption associated with metadata access operations.

Inventive Principle:
Principle #24Intermediary (Mediator)

4Quantity of substance

If compression-decompression engine is implemented, then bandwidth requirements and power consumption are reduced, but device complexity increases

Engineering Contradiction:
Improvebandwidth requirementsVSAvoiddevice complexity
Core Design Contradiction:
Quantity of substanceVSDevice complexity

Solution Approach 1:

The patent applies compression algorithms that transform cryptographic metadata into a more compact representation, changing the parameter of data size. This parameter change reduces the bandwidth requirements and power consumption associated with metadata transfers, while the added complexity of compression/decompression logic is offset by the significant resource savings.

Inventive Principle:
Principle #35Parameter changes

Data Source

PatentEP3391226B1Techniques to compress cryptographic metadata for memory encryption
Publication Date: 2021.06.23 INTEL CORP
  • EP3391226B1 patent drawingFigure 1
  • EP3391226B1 patent drawingFigure 2
  • EP3391226B1 patent drawingFigure 3

AI summary

Examples include techniques for compressing counter values included in cryptographic metadata. In some examples, a cache line to fill a cache included in on-die processor memory may be received. The cache arranged to store cryptographic metadata. The cache line includes a counter value generated by a counter. The counter value to serve as version information for a memory encryption scheme to write a data cache line to a memory location of an off-die memory. In some examples, the counter value is compressed based on whether the counter value includes a pattern that matches a given pattern and is then stored to the cache. In some examples, a compression aware and last recently used (LRU) scheme is used to determine whether to evict cryptographic metadata from the cache.