<HTML>
<HEAD>
<TITLE>XFS - File System Technical Specifications</TITLE>
</HEAD>
<BODY>
<IMG SRC="/toolbox/www/images/sgilogo.gif" WIDTH="130" HEIGHT="108" ALT="SGI" ALIGN="TOP"><BR><P>
<TABLE BORDER=1 CELLPADDING=5><CAPTION>
<CENTER><H2 ALIGN="CENTER"></H2>
</CENTER><CENTER><P ALIGN="CENTER"><IMG SRC="Technical.gif" ALT="Technical Specifications" WIDTH="476" HEIGHT="44">
</CENTER><P></CAPTION><TR VALIGN=TOP><TD WIDTH=50% ROWSPAN=2><FONT SIZE="-1">
<P><B>XFS File System</B>
<DL>
<DT>Technology
<DD>Journaled 64-bit file system for IRIX(tm), with guaranteed file system consistency
<DT>Product Span
<DD>Available as a layered product on all systems which run IRIX 5.3 or later, except IP4 and IP6 platforms
<DT>Max. File Size
<DD>1TB at first release, future releases support 9 million TB or the system drive limits
<DT>Max. File System Size
<DD>1TB at first release, future releases support 18 million TB or the system drive limits
<DT>File System Block Size
<DD>Selectable at file system creation time using mkfs_512, 512 bytes to 64KB for normal data, and up to 1MB for real-time data
<P>File system extents (contiguous data) are configurableat file creation time using <i>fcntl</i> and are multiples of the file system block size
<DT>Physical Disk Sector Size Supported 512 Bytes
<DD>NFS Compatibility Using NFS 5.3, 64-bit file systems can be exported to other systems which support the NFS V3 Protocol Systems which use NFS V2 protocol may access XFS file systems within the 32-bit limit imposed by the protocol
<DT>EFS Interoperability
<DD>EFS and XFS filesystems can be active on a single computer system
<DT>Backup/Restore
<DD>Dump/restore, bru, cpio, tar; IRIS NetWorker(tm) for IRIX
<DT>Dump Interchangeability
<DD>Restore can restore EFS dumps to either EFS or XFS file systems xfs_restore can restore XFS dumps to either XFS or EFS file systems
<DD>EFS file systems are dumped in EFS format, XFS file systems are dumped in XFS format
<DD>Dumps of active XFS file systems are possible
<DT>Support for Hierarchical Storage
<DD>Data Management API (DMIG) allows implementation of hierachial storage management software with no kernal modifications as well as high-performance dump programs without requiring "raw" access to the disk and knowledge of filesystem structures.
<DT>Swap to Files
<DD>Swap to files is supported
<DT>Guaranteed Rate I/O <i>*</i>
<DD>Hard and soft guarantees are supported, hard guarantees require turning off disk-drive self diagnostics Guarantees are expressed as a file descriptor, data rate, duration, and start time
<DT>Performance
<DD>Superior to EFS
<DT>Memory
<DD>32MB recommended
<DT>XFS Options
<DD>Guaranteed I/O for more than 4 streams
</DL>
</FONT></TD><TD WIDTH=50%><FONT SIZE="-1">
<P><B>XLV Volume Management</B>
<DL>
<DT>Topology Support
<DD>Mirroring, striping, concatenation supported.
<dd>Each volume can contain 3 subvolumes: real-time, log, and data
<dd>File system journal can be stored on a separate partition for maximum performance
<dd>Separate real-time partition for XFSfilesystems.
<DT>Maximum Disks per File System
<DD>Each sub volume can be independently mirrored up to 4 copies, striped up to 128 ways, and concatenated up to 100 volume extents (note: there may be hardware imposed limits)
<DT>Root Partition
<DD>Root partitions can be mirrored. Striping and device concatenation of root partitions is not supported at first release /usr partition mirroring, striping, and concatenationare available for the /usr partitions as for any other (non-root) file system
<DT>Dynamic Change
<DD>Add or remove a mirror, grow a volume, replace failed elements of a mirrored volume
<DT>XLV Options <i>*</i>
<DD>File system mirroring. Striping and concatenation are provided with base software
</DL>
</FONT></TD></TR>
<TR><TD NOBORDER><FONT SIZE="-1"><i>*</i> denotes unbundled product</FONT></TD></TR>
</TABLE>
<p>
#copyright 1995
</BODY>
</HTML>