4results about How to "Reduce administrative overhead" patented technology

This invention provides a queue management and invocation method, a network interface card (NIC) device, and a storage medium. The queue management and scheduling method is applied to a NIC device for inter-cloud service business interaction. The method includes: caching a first data packet for the inter-cloud service business interaction; performing an enqueue operation on the enqueued information packet from a queue management linked list of managed packet cache addresses, based on configuration information at a first scheduling granularity and configuration information at a second scheduling granularity, where the queue management linked list includes a local queue linked list and a virtual queue linked list; and performing dequeueing and scheduling of the local queue at the packet cache address after the enqueued information packet, based on multi-level scheduling, where the multi-level scheduling includes configurations at the first and second scheduling granularities. This invention reduces queue management overhead, improves queue management efficiency, and enables flexible queue scheduling at a smaller granularity.

This invention discloses a method, system, storage medium, and electronic device for unloading KV Cache during the LLM decoding phase. The method includes: writing the KV Cache data items to be unloaded to a ZNS SSD; maintaining the mapping relationship in a globally extended hash index structure; obtaining the current valid physical location information of the target KV Cache data; and performing garbage collection on zones in the ZNS SSD that meet preset reclamation conditions. The system includes: a data writing module, an index maintenance module, a query readback module, and a garbage collection module. This avoids forcibly unloading the KV Cache to the SSD during the computationally intensive pre-filling phase, focusing on optimizing the KV Cache capacity expansion and on-demand readback process during the decoding phase. It reduces index maintenance overhead and device write management overhead, improving cache management efficiency in long context and high-concurrency inference scenarios.

The application provides an efficient matrix storage method and system based on wind power prediction data, and relates to the technical field of wind power data storage. The method comprises the following steps: constructing multi-dimensional time series data according to wind power prediction data of each wind turbine in a wind farm; constructing a table structure based on an array field for the multi-dimensional time series data, each row corresponding to prediction data of a prediction model of the wind turbine at a prediction release time, and each feature storing the numerical value of the corresponding feature at multiple prediction target times in the form of an array field; performing hierarchical storage management on the wind power prediction data according to the table structure, storing wind power prediction data within a first time range as hot data in a cache table in the form of an original array; and storing the remaining wind power prediction data as cold data, obtaining multiple time blocks, performing singular value decomposition on the original matrix corresponding to each time block, and compressing and storing the cold data into a cold data table. The application helps to improve data calling efficiency and reduce overall storage space occupation.

The invention discloses a storage main control chip buffer area management method and system, and relates to the technical field of storage. The buffer area comprises the main buffer area and the auxiliary buffer area, real-time and accurate monitoring of the space use state of the buffer area is achieved through the register, an auxiliary buffer area power-on triggering mechanism based on the residual space threshold value is assisted, the auxiliary buffer area is dynamically awakened only when the space of the main buffer area is about to be exhausted, and the time of the main buffer area is shortened. On one hand, the auxiliary buffer is prevented from being in a no-load power supply state for a long time, and on the other hand, the data is preferentially and intensively stored in the main buffer, and data integrity and system response speed are ensured while power consumption management is realized.