man/man8/venti.8

.TH VENTI 8
.SH NAME
venti \- an archival block storage server
.SH SYNOPSIS
.B venti/venti
[
.B -dsw
]
[
.B -a
.I ventiaddress
]
[
.B -B
.I blockcachesize
]
[
.B -c
.I config
]
[
.B -C
.I cachesize
]
[
.B -h
.I httpaddress
]
[
.B -I
.I icachesize
]
.PP
.B venti/sync
[
.B -h
.I host
]
.SH DESCRIPTION
.I Venti
is a block storage server intended for archival data.
In a Venti server,
the SHA1 hash of a block's contents acts as the block
identifier for read and write operations.
This approach enforces a write-once policy, preventing accidental or
malicious destruction of data.  In addition, duplicate copies of a
block are coalesced, reducing the consumption of storage and
simplifying the implementation of clients.
.PP
Storage for
.I venti
consists of a data log and an index, both of which
can be spread across multiple files.
The files containing the data log are themselves divided into self-contained sections called arenas.
Each arena contains a large number of data blocks and is sized to
facilitate operations such as copying to removable media.
The index provides a mapping between the a Sha1 fingerprint and
the location of the corresponding block in the data log.
.PP
The index and data log are typically stored on raw disk partitions.
To improve the robustness, the data log should be stored on
a device that provides RAID functionality.  The index does
not require such protection, since if necessary, it can
can be regenerated from the data log.
The performance of
.I venti
is typically limited to the random access performance
of the index.  This performance can be improved by spreading the
index accross multiple disks.  
.PP
The storage for
.I venti
is initialized using
.IR fmtarenas ,
.IR fmtisect ,
and
.I fmtindex
(see
.IR ventiaux (8)).
A configuration file,
.IR venti.conf (6),
ties the index sections and data arenas together.
.PP
A Venti
server is accessed via an undocumented network protocol.
Two client applications are included in this distribution:
.IR vac (1)
and
.IR vacfs (4).
.I Vac
copies files from a Plan 9 file system to Venti, creating an
archive and returning the fingerprint of the root.
This archive can be mounted in Plan 9 using 
.IR vacfs .
These two commands enable a rudimentary backup system.
A future release will include a Plan 9 file system that uses
Venti as a replacement for the WORM device of 
.IR fs (4).
.PP
The
.I venti
server provides rudimentary status information via
a built-in http server.  The URL files it serves are:
.TP
.B stats
Various internal statistics.
.TP
.B index
An enumeration of the index sections and all non empty arenas, including various statistics.
.TP
.B storage
A summary of the state of the data log.
.TP
.B xindex
An enumeration of the index sections and all non empty arenas, in XML format.
.PP
Several auxiliary utilities (see
.IR ventiaux (8))
aid in maintaining the storage for Venti.
With the exception of
.I rdarena ,
these utilities should generally be run after killing the
.I venti
server.
The utilities are:
.TP
.I checkarenas
Check the integrity, and optionally fix, Venti arenas.
.TP
.I checkindex
Check the integrity, and optionally fix, a Venti index.
.TP
.I buildindex
Rebuild a Venti index from scratch.
.TP
.I rdarena
Extract a Venti arena and write to standard output.
.PD
.PP
Options to 
.I venti
are:
.TP
.BI -a " ventiaddress
The network address on which the server listens for incoming connections.
The default is
.LR tcp!*!venti .
.TP
.BI -B " blockcachesize
The size, in bytes, of memory allocated to caching raw disk blocks.
.TP
.BI -c " config
Specifies the
Venti
configuration file.
Defaults to
.LR venti.conf .
.TP
.BI -C " cachesize
The size, in bytes, of memory allocated to caching 
Venti
blocks.
.TP
.BI -d
Produce various debugging information on standard error.
.TP
.BI -h " httpaddress
The network address of Venti's built-in
http
server.
The default is
.LR tcp!*!http .
.TP
.BI -I " icachesize
The size, in bytes, of memory allocated to caching the index mapping fingerprints
to locations in 
.IR venti 's
data log.
.TP
.B -s
Do not run in the background.
Normally,
the foreground process will exit once the Venti server
is initialized and ready for connections.
.TP
.B -w
Enable write buffering.  This option increase the performance of writes to
.I venti
at the cost of returning success to the client application before the
data has been written to disk.
The server implements a
.I sync
rpc that waits for completion of all the writes buffered at the time
the rpc was received.
Applications such as
.IR vac (1)
and the
.I sync
command described below
use this rpc to make sure that the data is correctly written to disk.
Use of this option is recommended.
.PD
.PP
The units for the various cache sizes above can be specified by appending a
.LR k ,
.LR m ,
or
.LR g
to indicate kilobytes, megabytes, or gigabytes respectively.
The command line options override options found in the
.IR venti.conf (6)
file.
.PP
.I Sync
connects to a running Venti server and executes a sync rpc
(described with the
.B -w
option above). 
If sync exits successfully, it means that all writes buffered at the
time the command was issued are now on disk.
.SH SOURCE
.B /sys/src/cmd/venti
.SH "SEE ALSO"
.IR venti.conf (6),
.IR ventiaux (8),
.IR vac (1),
.IR vacfs (4).
.br
Sean Quinlan and Sean Dorward,
``Venti: a new approach to archival storage'',
.I "Usenix Conference on File and Storage Technologies" ,
2002.