Hierarchical Storage Management
The CloudTier Storage Tiering Filter Driver SDK is a Hierarchical Storage Management (HSM) development kit. It provides a kernel-level framework to implement transparent data migration between high-cost (hot) and low-cost (cold) storage media. By replacing migrated files with Stub Files, it enables organizations to optimize storage efficiency and reduce OpEx/CapEx without disrupting existing applications.
System Architecture
The SDK operates at the file system filter level, acting as an "invisible traffic cop" between user applications and the Windows NTFS/ReFS file system.
Kernel-Mode Filter Driver: Intercepts I/O requests (IRP_MJ_READ, IRP_MJ_CREATE) and manages reparse points. It evaluates requests against business rules to determine if data resides locally or requires a remote fetch.
User-Mode Service (Policy & Logic): Manages cloud connectivity (S3, Azure, etc.) and executes the Lifecycle Management policies.
Stub Management: Leverages Windows reparse points and sparse files. Stubs retain full original metadata (timestamps, ACLs, file size) while consuming minimal local disk space.
Understanding the Stub File
A core component of the CloudTier architecture is the Stub File, which acts as a seamless placeholder for data that has been migrated to the cloud or secondary storage.
Technical Composition
Technically, a stub file is a sparse file tagged with a reparse point attribute. This reparse point acts as a trigger for the CloudTier kernel-level filter driver. When an application attempts to open or read the stub, the OS recognizes the reparse point and hands control to the filter driver, which then fetches the requested data from the remote tier.
Metadata and Custom Data
A stub file mimics a standard physical file in appearance and behavior. It retains 100% of the original file's metadata, including:
File name and extension
Logical file size (applications see the full original size, not the size on disk)
Creation time, last write time, and last access time
Security attributes and Access Control Lists (ACLs)
Additionally, stub files can store custom tag data embedded directly within the stub metadata, allowing applications to track internal database IDs, migration job details, or routing URIs.
The User Experience
Unlike physical files, a stub file consumes virtually zero local disk space because it does not store the actual file content. However, to end-users and applications, the file appears entirely normal. Because the file remains visible in Windows Explorer and command-line tools, CloudTier prevents the "broken link" syndrome common in legacy archiving software.
Comments
Post a Comment