How DoubleSpace Uses Sector Allocation (98407)

The information in this article applies to:
    Microsoft MS-DOS operating system 6.0
    Microsoft MS-DOS operating system 6.2
    Microsoft MS-DOS operating system 6.22
    Microsoft Plus! for Windows 95
    Microsoft Windows 95

This article was previously published under Q98407
This information applies to both Microsoft DoubleSpace and Microsoft DriveSpace. For MS-DOS 6.00 and 6.20, use DBLSPACE in place of DRVSPACE for commands and filenames.


DriveSpace uses sector allocation granularity even though you cannot see evidence of this in the Bytes Available On Disk value reported by CHKDSK.


Since the file allocation table (FAT) system is cluster based, DriveSpace uses its own FAT (which is called the MDFAT [Microsoft DriveSpace FAT]) to achieve sector allocation. Each entry in the FAT corresponds to an entry in the MDFAT, as well as a logical cluster of 8 kilobytes (K) for MS-DOS and Windows 95 DriveSpace, or 32 K for Plus! DriveSpace 3. Each MDFAT entry also corresponds to a sequence of sectors less than or equal to 8K (MS-DOS and Windows 95 DriveSpace) or 32 K (Plus! DriveSpace 3).

NOTE: Plus! DriveSpace 3 supports fragmented MDFAT clusters, so a 32 K DriveSpace 3 MDFAT cluster need not necessarily be contiguous.

Since disk size reporting is accomplished by adding up the total number of clusters and then multiplying by the cluster size, there is no way to get a non-cluster size value for Total Disk Space or Bytes Available On Disk values from CHKDSK.

Estimated Compression Ratio (ECR)

When you run CHKDSK on a DriveSpace drive, it lists the number of used clusters in terms of uncompressed data. When it lists the number of unused clusters, it must specify some multiple of clusters based on how many compressed clusters it thinks can fit inside the actual free space. It does this by taking the ECR and multiplying it by the actual free space on the drive (in sectors) and then dividing that by the number of sectors in a cluster (16).

As a result, free disk space can only be specified as a integral number of clusters.

Modification Type: Major Last Reviewed: 12/17/2000
Keywords: KB98407