US 10,949,387 B1 · Scalable filesystem enumeration and metadata operations

Abstract

Systems, apparatus, and methods are disclosed for performing scalable operations in a file system, including POSIX-like file systems. Metadata entries in a namespace or directory tree are sharded across multiple file metadata servers. An enumeration operation, such as listing a directory, is parallelized across the multiple file metadata servers, while retaining standard functionality transparently to clients. Other enumeration operations include no-output operations such as changing file attributes or deleting a file, and cumulative operations such as counting total disk space usage. The parallelization is compatible with tree-level parallelization and storage-level parallelization. Disclosed technologies can be applied to other fields requiring scalable enumeration, such as database and network applications.

Claims19

Dependent claims

A method, comprising:

by a computer:
receiving, from a client, a first command to enumerate objects stored in the file system, and, responsive to the first command:
initializing a plurality of data structures, each of the data structures being associated with a respective one of a plurality of metadata servers of a file system;
and
invoking a respective parallel enumeration task at each of the plurality of metadata servers, wherein the metadata servers maintain respective non-overlapping collections of metadata of the objects stored in the file system;
and
by at least two of the plurality of metadata servers:
using the respective data structures, traversing respective metadata entries, each of the metadata entries representing a corresponding target object of the file system;
and
performing an enumeration function with each traversed metadata entry.

7 more elements

A computer-implemented system comprising:

a plurality of metadata servers, comprising a primary metadata server and two or more secondary metadata servers, each of the plurality of metadata servers comprising one or more processors, memory coupled thereto, and one or more network adapters;
a first metadata store directly controlled by the primary metadata server and configured to store metadata entries for directories of a file system;
a plurality of second metadata stores directly controlled by respective ones of the secondary metadata servers and configured to store metadata entries for files of the file system, wherein the files are organized within the directories, each of the second metadata stores is distinct from the first metadata stores, so that the metadata entries for the directories are stored in a distinct store from the metadata entries for the files;
and
wherein the plurality of metadata servers are interconnected by the network adapters and are configured to perform a method, the method comprising:
by the primary metadata server:
responsive to receiving a first command to add a first metadata entry for a directory at a first target location in a file system:
determining an inode value for the first metadata entry;
constructing the first metadata entry for the directory;
and
storing the first metadata entry in the first metadata store;
and
responsive to receiving a second command to add a second metadata entry for a file at a second target location in the file system:
associating the second metadata entry with a first one of the secondary metadata servers;
and
sending a communication to the first secondary metadata server related to the second metadata entry;
and
by the first secondary metadata server, responsive to the communication:
determining an inode value for the second metadata entry;
and
storing the second metadata entry in a corresponding one of the second metadata stores directly controlled by the first secondary metadata server.

18 more elements

Dependencies are inferred from the claim text.

Scalable filesystem enumeration and metadata operations

Life of this patent

Scalable filesystem enumeration and metadata operations

Life of this patent

Abstract

Claims19