Electrical computers and digital processing systems: memory – Storage accessing and control – Hierarchical memories
Patent
1996-08-07
1998-07-14
Asta, Frank J.
Electrical computers and digital processing systems: memory
Storage accessing and control
Hierarchical memories
711113, 711119, 711143, 711147, 39520057, 39520062, 39520033, G01R 3128, G06F 1100
Patent
active
057817332
DESCRIPTION:
BRIEF SUMMARY
FIELD OF THE INVENTION
The present invention relates to data storage in computer networks, and more specifically to a method and apparatus for combining writes and avoiding redundant writes when flushing cached data across a network.
TECHNICAL BACKGROUND OF THE INVENTION
On a typical computer, the central processor is separated from a non-volatile data storage medium by a bus, network, or other communication channel. For instance, in many networks modified data must be transferred across a network link before it can be written on non-volatile storage such as a hard disk.
Personal computers and workstations are often linked in a computer network to promote the sharing of data, application programs, files, and other resources. Sharing is accomplished with the use of a network "server." The server is a processing unit dedicated to managing centralized resources, managing data, and sharing resources with the personal computers and workstations, which are referred to as "clients" of the server.
Unfortunately, communicating over a network can be time-consuming. In many client/server networks, the clients, the server, or both therefore retain copies of information in a local "cache" that is readily accessed. Cache memory is divided into regions known as "cache blocks." Blocks that have been written to but not yet copied to non-volatile storage are called "dirty" blocks.
A drawback of conventional systems is their performance of redundant write operations. In some cases, entire blocks are written across the network even though only a portion of the block was overwritten in cache. Thus, much "clean" data is unnecessarily transferred through one or more communication channels to the server's hard disk. Moreover, conventional systems assume that the data values stored in a cache block are changed whenever data is written to that cache block. In fact, the previously cached values do not always differ from the values written over them. As a result, the same data may be sent through the communication channel twice during consecutive writes, to be stored again at the same location on the server's disk. Such transfers add to network congestion without preserving new data.
Another drawback of conventional systems is their inefficient use of available network bandwidth. The relationship between network packet size, current round trip time, and numerous other factors varies with time. Under some conditions it is more efficient to send one large packet containing dirty data, and under other conditions it is more efficient to send two separate smaller packets. Conventional systems for flushing cached data to a server do not adequately reflect these factors.
Thus, it would be an advancement in the art to provide an apparatus and method for decreasing the quantity of redundant write operations in a computer network.
It would also be an advancement to provide such an apparatus and method which dynamically adjust to changing network conditions to use one larger packet or two smaller packets, as appropriate.
Such an apparatus and method for redundant write removal are disclosed and claimed herein.
BRIEF SUMMARY OF THE INVENTION
The present invention provides an apparatus and method for selecting cached data to write across a computer network from a client to a server. The client includes a client data cache, which may be either a conventional cache or a cache according to the '919 application. One method includes the computer-implemented step of determining a write-bridge size, in a manner set forth after use of the write-bridge is described.
Additional steps locate first and second dirty regions in the client data cache. The dirty regions are separated from one another by a clean region of a determinable size. The dirty regions may be defined in alternative ways according to the present invention. Under one definition, a byte of cache is dirty if it was overwritten after the most recent flushing of data stored at that location. Under an alternative definition, the byte is dirty only if it has been thus. overwritten and if t
REFERENCES:
patent: 4858111 (1989-08-01), Steps
patent: 5119485 (1992-06-01), Ledbetter, Jr. et al.
patent: 5133519 (1992-05-01), Johnson et al.
patent: 5155824 (1992-10-01), Edenfield et al.
patent: 5283868 (1994-02-01), Baker et al.
patent: 5377338 (1994-12-01), Olson et al.
patent: 5408644 (1995-04-01), Schneider et al.
patent: 5452447 (1995-09-01), Nelson et al.
patent: 5542066 (1996-07-01), Mattson et al.
patent: 5668967 (1997-09-01), Olson et al.
Asta Frank J.
Novell Inc.
Pierce Ivan
LandOfFree
Apparatus and method for redundant write removal does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Apparatus and method for redundant write removal, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Apparatus and method for redundant write removal will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-1892270