A metadata object access method, apparatus, device and medium

Abstract

The present application discloses a method, device, device and medium for accessing a metadata object. The method includes: determining a target object ID corresponding to the metadata object access request according to the obtained metadata object access request; The target object ID is processed to obtain the target metadata object ID corresponding to the metadata object access request; and the target metadata object corresponding to the target metadata object ID in the metadata object set is accessed. In this way, when multiple metadata object access requests are received and the corresponding object IDs are in the same volume, the corresponding target metadata object can also be determined according to the object ID, so that the metadata operations of the same volume can be distributed to different metadata objects. On data objects, the concurrency of metadata access is improved, the occurrence of hot metadata objects is avoided, and the pressure on each metadata object is reduced, thereby improving volume performance and overall data IO performance.

Description

technical field [0001] The present application relates to the technical field of distributed storage, and in particular, to a method, apparatus, device and medium for accessing metadata objects. Background technique [0002] Distributed block storage provides block storage services to applications through the corresponding block storage library interface in the distributed storage system, and mainly provides virtual disks for virtual machines on cloud platforms. Distributed block storage is similar to traditional SAN (StorageArea Network and SAN Protocol, storage area network) storage, and both provide block-level access. In distributed block storage, every time a volume is created, a metadata object (object-map object) will be created correspondingly, and the bitmap (bitmap information) of the "object" of the volume will be recorded through the metadata object, so that it can be relatively fast. Some management functions of the processing block, such as: fast calculation o...

AI Technical Summary

Problems solved by technology

However, while the metadata object provides the fast IO function of the block, it also causes the performance bottleneck of IO due to its own reasons.

For a newly created volume, when creating a write object, in order to ensure the data consistency of the distributed block storage records, during the IO operation, you must first ensure that the metadata object of the volume is updated before performing the data IO operation. The write IO on a single volume is waiting for the write operation on the metadata object, resulting in hot metadata objects, resulting in too heavy a burden on the metadata object, resulting in a decrease in data IO performance, and the larger the created volume, the newer The greater the impact caused by the volume write operation

Images