Large Data Read Prefetch for Low-Latency Storage Transfer
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional memory systems face inefficiencies in handling large data transfers due to limitations in atomic unit read commands, leading to substantial latency and power usage for sequential reads, especially in mobile devices with constrained cache sizes.
Innovation Solution
Implementing a host-resident logical-to-physical mapping (L2P) cache and enhanced read commands that allow for simultaneous transfer of multiple atomic chunks, eliminating the need for repetitive single-unit read commands.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If single atomic unit read commands are used, then the storage system can handle individual read requests reliably, but the latency and power consumption increase substantially for large data transfers
Solution Approach 1:
The patent segments large data transfer operations into multiple atomic unit reads that can be executed in parallel. The host device divides a large read request into multiple smaller sub-requests, each targeting a specific atomic unit (page or block) in the storage device. These segmented requests are then processed simultaneously through multiple I/O channels, resolving the contradiction by maintaining atomic unit reliability while achieving efficient large data transfer through parallel execution of segmented operations.
Solution Approach 2:
The patent implements preliminary action by pre-fetching data blocks into the storage device's cache memory before the host actually needs them. The storage device proactively reads data blocks that are likely to be accessed soon and stores them in cache, so when the host issues read commands, the data is already available or can be quickly retrieved. This eliminates wait time and reduces latency for large data transfers while maintaining the atomic unit read command structure.
2Reliability
If multiple single-unit read commands are issued sequentially, then each read request can be processed accurately, but the total power consumption increases for large data transfers
Solution Approach 1:
The patent segments the sequential read process into parallel operations across multiple I/O channels. Instead of issuing multiple single-unit read commands sequentially through one channel, the host divides the read request into segments and sends them simultaneously through multiple independent channels. This segmentation maintains data transfer accuracy for each atomic unit while dramatically reducing total power consumption by eliminating the sequential execution overhead and utilizing parallel processing capabilities of the storage device.
Solution Approach 2:
The patent merges multiple atomic unit read operations into a single coordinated parallel execution framework. Multiple I/O channels that would otherwise operate independently are merged into a unified parallel read operation, where multiple atomic units are read simultaneously through different channels. This combining approach maintains the reliability of individual atomic reads while reducing overall power consumption by executing operations in parallel rather than sequentially, leveraging the storage device's ability to handle multiple concurrent operations.
3Volume of moving object
If the cache size is constrained as in mobile devices, then the device portability is improved, but the ability to buffer large data transfers is reduced
Solution Approach 1:
The patent segments large data transfer operations into multiple atomic unit reads that can be executed in parallel through multiple I/O channels. This segmentation allows the system to bypass the limitation of small cache size by processing data in smaller atomic units that fit within the constrained cache, while achieving large data transfer efficiency through parallel execution. The host device divides large read requests into segments, each manageable within the limited cache capacity, and processes them simultaneously rather than sequentially.
Solution Approach 2:
The patent implements preliminary action by pre-fetching data blocks into the storage device's limited cache memory before the host needs them. The storage device proactively reads and caches data blocks that are likely to be accessed soon, maximizing the utilization of the constrained cache space. This pre-fetching mechanism allows mobile devices with small caches to achieve high productivity for large data transfers by having data ready in cache before actual read operations, eliminating wait time and improving overall transfer efficiency despite the small cache volume.
Data Source
Figure 1
Figure 2
Figure 3A~3B
AI summary
Devices and techniques are disclosed herein for more efficiently exchanging large amounts of data between a host and a storage system. In an example, a large read operation can include receiving a pre-fetch command, a parameter list and a read command at a storage system. In certain examples, the pre-fetch command can provide an indication of the length of the parameter list, and the parameter list can provide location identifiers of the storage system from which the read command can sense the read data.