Parallel file system and buffer management arbitration

• US 5,963,963 A
• Filed: 07/11/1997
• Issued: 10/05/1999
• Est. Priority Date: 07/11/1997
• Status: Expired due to Term

• Current Assignee: Google LLC (Alphabet Inc.)
• Original Assignee: International Business Machines Corporation
• Inventors: Schmuck, Frank B., Haskin, Roger Lee, Wyllie, James Christopher, McNabb, Daniel Lloyd
• Primary Examiner(s): Lintz, Paul R.
• Assistant Examiner(s): Colbert, Ella
• Application Number:
  US08/893,722
• Time in Patent Office: 816 Days
• Field of Search: 707/205, 707/10, 711/114
• US Class Current: 1/1
• CPC Class Codes: G06F 16/1858 Parallel file systems, i.e. file systems supporting multiple processors
• CPC Class Codes: G06F 2209/522 Manager
• CPC Class Codes: G06F 9/5016 the resource being the memory
• CPC Class Codes: G06F 9/526 Mutual exclusion algorithms
• CPC Class Codes: Y10S 707/99956 File allocation

Abstract

A computer system having a shared disk file system running on multiple computers each having their own instance of an operating system and being coupled for parallel data sharing access to files residing on network attached shared disks. Methods are provided for use as a parallel file system in a shared disk environment by use of a scalable directory service for the system with a stable cursor, and a segmented allocation map. Dynamic prefetch and cached balance pools for multiple accesses improve the system. Extended file attributes are used for implementation of Access Control Lists in a parallel file system. Improvements to caching and cache performance developments balance pools for multiple accesses. A metadata node manages file metadata for parallel read and write actions. For managing data for the same file distributed across multiple disk devices, a buffer manager is provided for a file system buffer pool which arbitrates use of memory resources among different system components competing for memory with buffer information for a component that said buffer manager needs for deciding how much memory to allocate among components of the system. The buffer information includes a desired memory size and current activity level. The buffer manager provides to a system component data as to how much memory has been assigned for use by that component of the system, and it schedules account resources to maximize file system throughput.

Claims

1. In a computer system employing a file system for storing file data and allowing multiple nodes of said computer system to allocate space for storing file data simultaneously in a shared-disck file system, a method comprising:

providing a file system in which data for the same file is distributed across multiple disk devices, and

providing a buffer manager for a file system buffer pool which arbitrates use of memory resources among different system components competing for memory with buffer information for a component that said buffer manager needs for deciding how much memory to allocate among components of the system, said buffer information including a desired memory size and current activity level and then

providing from said buffer manager to said component data as to how much memory has been assigned for use by that component of the system.

2. In a computer system method according to claim 1, in which one of the system components requiring assignment of memory resources for use is a file system buffer pool which is provided for caching recently accessed file data and data that was prefetched for sequential readers and said providing of information from said components includes providing

said buffer manager with appropriate information to take into account resources for prefetching and for

scheduling the resources assigned by the buffer manager so as to maximize file system throughput.

3. In a computer system method according to claim 1, wherein said file system buffer pool is logically divided into a pool part used for prefetching and a general pool part used for caching recently accessed file blocks.

4. In a computer system method according to claim 3, wherein said pool parts are presented to said buffer manager as two separate components for the file system to compute separate desired memory sizes and activity levels for the pool parts.

5. In a computer system method according to claim 4, wherein said activity levels for the pool parts are computed by counts measuring data access rates for each pool, and wherein on each buffer access the appropriate count is updated based on whether the buffer is accessed by sequential or random I/O.

6. In a computer system method according to claim 3, wherein said general pool is computed by measuring working sets using reference bits and counters to determine the total amount of distinct file data accessed over some time period.

7. In a computer system method according to claim 3, wherein the size of the prefetch pool is computed by determining the number and capability of disk drives belonging to the file system and as well determining the number of files being accessed sequentially, and as well determining the rate at which data is being read.

8. In a computer system method according to claim 3, wherein the method of computing the prefetch pool for multiple application programs and multiple disks per file system, computing a number of prefetch buffers required so that:

(a) if the combined data rate at which all, the application programs attempt to read data is less than the total available disk bandwidth, then data will be supplied to each application as fast as it reads the data, with no I/O waits, but (b) if the combined data rate of the application programs is greater than the total available disk bandwidth, then data will be read as fast as it can be retrieved from disk.

9. In a computer system method according to claim 3, wherein is included:

scheduling prefetch I/O by using an actual number of prefetch buffers assigned by said buffer manager,

maintaining two global counters for a first number of prefetch I/Os currently in progress and a second number of buffers holding blocks to be read and summing said first number and said second number to determine the number of buffers currently in use for prefetching, and tracking the progress of the numbers and updating the pool assignments by said buffer manager.