Browse Prior Art Database

Adaptive Data Staging Mechanism in a Virtual Storage System

IP.com Disclosure Number: IPCOM000081519D
Original Publication Date: 1974-Jun-01
Included in the Prior Art Database: 2005-Feb-28
Document File: 6 page(s) / 34K

Publishing Venue

IBM

Related People

Clark, NR: AUTHOR [+2]

Abstract

In a storage system consisting of two levels of storage (one being faster than the other), it is desired that staging, data transfers between levels, be held to a minimum. This description proposes a method by which the system itself could optimize the amount of data staged. Key to the proposal is the maintenance of usage statistics associated with each page of data on the low-level storage media. The use of this scheme allows the storage system to adapt its staging for optimum performance based on current page usage.

This text was extracted from a PDF file.
At least one non-text object (such as an image or picture) has been suppressed.
This is the abbreviated version, containing approximately 27% of the total text.

Page 1 of 6

Adaptive Data Staging Mechanism in a Virtual Storage System

In a storage system consisting of two levels of storage (one being faster than the other), it is desired that staging, data transfers between levels, be held to a minimum. This description proposes a method by which the system itself could optimize the amount of data staged. Key to the proposal is the maintenance of usage statistics associated with each page of data on the low-level storage media. The use of this scheme allows the storage system to adapt its staging for optimum performance based on current page usage.

For the purpose of this description assume that a storage system consisting of two levels of storage are provided -- high and low levels (see the figure). The high level is significantly faster than the low level. Accessing the storage system are n number of hosts, central processing units, where n may be any integral number. When a host anticipates accessing certain data sets, an Acquire command is given to the storage system which specifies the data to be staged from the low-level storage to the high-level storage, by specifying a volume ID and the data extents. Some definitions are: Volume: The basic unit of storage within the low-level

storage.

Data Set: A related group of data which is maintained on a

volume. More than one data set may be maintained on

a given volume.

Data Extent: The upper and lower boundaries within a volume which

contain the data to be acted on by an "Acquire

command".

Page: The basic unit of storage which is allocated

within

the high-level storage to contain data as a result

of an Acquire command. A given data extent may

include several pages and partial pages at each

boundary. That is, a data extent may start in the

middle of a page and end in the middle of a page.

Cylinders: The addressable boundaries within a page are known as data cylinders.

At the time an Acquire command is received by the storage system, storage within the high-level storage is allocated so that each page affected can be staged into the high-level store. The question now is, which pages should be staged? There are several alternatives. 1) Stage all pages immediately.

2) Stage no page until an access of data within that page is

made. This is called "page faulting".

3) Have the Acquire command define which pages to stage.

4) Use an adaptive procedure which will stage only those

pages which are likely to be accessed,

based on past history. The history may be constantly

updated and maintained at a specific location

1

Page 2 of 6

for that page on the associated volume in the

low-level storage.

It is method number 4 that is being described here. First, a brief discussion of the other methods.
1. If all pages are staged immediately, then the maximum

amount of time is used in transferring

data from the low-level store to the high-level

store. Thus, time is wasted staging pages

which are not accessed as a result of this staging.
2. If no page is staged until data is ac...