<!DOCTYPE HTML PUBLIC "-//IETF//DTD HTML 3.0//EN" "http://www.w3.org/TR/html4/loose.dtd">
<HTML>
<HEAD><TITLE>
Hercules: Shared Device Server</TITLE>
<LINK REL=STYLESHEET TYPE="text/css" HREF="hercules.css">
</HEAD>
<BODY BGCOLOR="#ffffcc" TEXT="#000000" LINK="#0000A0"
VLINK="#008040" ALINK="#000000">
<h1>Hercules Shared Device Server</h1>
<hr noshade>
<h2>Contents</h2>
<ul>
<li><a href="#overview"> Overview </a>
<li><a href="#technical"> Technical BS </a>
<li><a href="#caching"> Caching </a>
<li><a href="#compression"> Compression </a>
<li><a href="#todo"> To Do </a>
</ul>
<p>
<hr noshade>
<h2><a NAME="overview">Overview</a></h2>
<p>
(C) Copyright Greg Smith, 2002-2007
<p>
Shared device support allows multiple Hercules instances to share
devices. The device will be 'local' to one instance and 'remote'
to all other instances. The local instance is the 'server' for
that device and the remote instance is the 'client'. You do not
have to IPL an operating system on the device server. Any number
of Hercules instances can act as a server in a "Hercplex".
<p>
To use a device on a remote system, instead of specifying a file
name on the device config statement, you specify
<p>
<code>ip_address_or_name:port:devnum</code>
<p>
For example:
<pre class="jcl">
0100 3350 localhost:3990:0100
</pre>
<p>
which says there is a device server on the local host listening
on port 3990 and we want to use its 0100 device as 0100. The
default port is 3990 and the default remote device number is the
local device number. So we could say:
<pre class="jcl">
0100 3350 localhost
</pre>
<p>
instead, providing we don't actually have a file 'localhost'.
Interestingly, the instance on the local host listening on 3990
could have a statement:
<pre class="jcl">
0100 3350 192.168.200.1::0200
</pre>
<p>
which means that instance in turn will use device 0200 on the
server at 192.168.200.1 listening on port 3990. The original
instance will have to 'hop' thru the second instance to get
to the real device.
<p>
Device sharing can be 'split' between multiple instances.
For example, suppose instance A has:
<pre class="jcl">
SHRDPORT 3990
0100 3350 localhost:3991
0101 3350 mvscat
</pre>
<p>
and instance B has:
<pre class="jcl">
SHRDPORT 3991
0100 3350 mvsres
0101 3350 localhost
</pre>
<p>
Then each instance acts as both a client and as a server.
<p>
When 'SHRDPORT' is specified, thread 'shared_server' is started
at the end of Hercules initialization. In the example above,
neither Hercules instance can initialize their devices until the
server is started on each system. In this case, the device trying
to access a server gets the 'connecting' bit set on in the DEVBLK
and the device still needs to initialize. After the shared server
is started, a thread is attached for each device that is connecting
to complete the connection (which is the device init handler).
<p><br>
<h2><a NAME="technical">Technical BS</a></h2>
<p>
There are (at least) two approaches to sharing devices. One is to
execute the channel program on the server system. The server will
need to request from the client system information such as the ccw
and the data to be written, and will need to send to the client
data that has been read and status information. The second is to
execute the channel program on the client system. Here the client
system makes requests to the server system to read and write data.
<p>
The second approach is currently implemented. The first approach
arguably emulates 'more correctly'. However, an advantage of the
implemented approach is that it is easier because only the
client sends requests and only the server sends responses.
<p>
Both client and server have a DEVBLK structure for the device.
Absurdly, perhaps, in originally designing an implementation for
shared devices it was not clear what type of process should be the
server. It was a quantum leap forward to realize that it could
just be another hercules instance.
<p>
<h3>Protocol</h3>
<p>
(If this section is as boring for you to read as it was for me
to write then please skip to the next section ;-)
<p>
The client sends an 8 byte request header and maybe some data:
<pre>
+-----+-----+-----+-----+-----+-----+-----+-----+
| cmd |flag | devnum | id | length |
+-----+-----+-----+-----+-----+-----+-----+-----+
<-------- length --------->
+----- . . . . . -----+
| data |
+----- . . . . . -----+
</pre>
<p>
'cmd' identifies the client request. The requests are:
<p>
<dl>
<dt>0xe0 CONNECT<p><dd>
Connect to the server. This requires
the server to allocate resources to
support the connection. Typically issued
during device initialization or after being
disconnected after a network error or timeout.
<p>
<dt>0xe1 DISCONNECT<p><dd>
Disconnect from the server. The server
can now release the allocated resources
for the connection. Typically issued during
device close or detach.
<p>
<dt>0xe2 START<p><dd>
Start a channel program on the device.
If the device is busy or reserved by
another system then wait until the device
is available unless the NOWAIT flag bit
is set, then return a BUSY code. Once
START succeeds then the device is unavailable
until the END request.
<p>
<dt>0xe3 END<p><dd>
Channel program has ended. Any waiters
for the device can now retry.
<p>
<dt>0xe4 RESUME<p><dd>
Similar to START except a suspended
channel program has resumed.
<p>
<dt>0xe5 SUSPEND<p><dd>
Similar to END except a channel program
has suspended itself. If the channel
program is not resumed then the END
request is <i>not</i> issued.
<p>
<dt>0xe6 RESERVE<p><dd>
Makes the device unavailable to any other
system until a RELEASE request is issued.
<i>Must</i> be issued within the scope of START/END.
<p>
<dt>0xe7 RELEASE<p><dd>
Makes the device available to other systems
after the next END request.
<i>Must</i> be issued within the scope of
START/END.
<p>
<dt>0xe8 READ<p><dd>
Read from a device. A 4-byte 'record'
identifier is specified in the request
data to identify what data to read in the
device context.
<i>Must</i> be issued within the scope of START/END.
<p>
<dt>0xe9 WRITE<p><dd>
Write to a device. A 2-byte 'offset' and
a 4-byte 'record' is specified in the request
data, followed by the data to be written.
'record' identifies what data is to be written
in the device context and 'offset' and 'length'
identify what to update in 'record'.
<i>Must</i> be issued within the scope of START/END.
<p>
<dt>0xea SENSE<p><dd>
Retrieves the sense information after an i/o
error has occurred on the server side. This
is typically issued within the scope of the
channel program having the error. Client side
sense or concurrent sense will then pick up the
sense data relevant to the i/o error.
<i>Must</i> be issued within the scope of START/END.
<p>
<dt>0xeb QUERY<p><dd>
Obtain device information, typically during
device initialization.
<p>
<dt>0xec COMPRESS<p><dd>
Negotiate compression parameters. Notifies the
server what compression algorithms are supported
by the client and whether or not data sent back
and forth from the client or server should be
compressed or not. Typically issued after CONNECT.
<p>
NOTE: This action should actually be SETOPT or
some such; it was just easier to code a COMPRESS
specific SETOPT (less code).
</dl>
<p><br>
'flag' qualifies the client request and varies by the request.
<dl>
<dt>0x80 NOWAIT<p><dd>
For START, if the device is unavailable then
return BUSY instead of waiting for the device.
<p>
<dt>0x40 QUERY<p><dd>
Identifies the QUERY request:
<p>
<dt>0x41 DEVCHAR<p><dd>
Device characteristics data
<p>
<dt>0x42 DEVID<p><dd>
Device identifier data
<p>
<dt>0x43 DEVUSED<p><dd>
Hi used track/block (for dasdcopy)
<p>
<dt>0x48 CKDCYLS<p><dd>
Number cylinders for CKD device
<p>
<dt>0x4c FBAORIGIN<p><dd>
Origin block for FBA
<p>
<dt>0x4d FBANUMBLK<p><dd>
Number of FBA blocks
<p>
<dt>0x4e FBABLKSIZ<p><dd>
Size of an FBA block
<p>
<dt>0x3x COMP<p><dd>
For WRITE, data is compressed at offset 'x':
<p>
<dt>0x2x BZIP2<p><dd>
using bzip2
<p>
<dt>0x1x LIBZ<p><dd>
using zlib
<p>
<dt>0xxy<p><dd>
For COMPRESS, identifies the compression
algorithms supported by the client (0x2y for bzip2,
0x1y for zlib, 0x3y for both) and the zlib compression
parameter 'y' for sending otherwise uncompressed data
back and forth. If 'y' is zero (default) then no
uncompressed data is compressed between client & server.
</dl>
<p><br>
'devnum' identifies the device by number on the server instance.
The device number may be different than the
device number on the client instance.
<p>
'id' identifies the client to the server. Each client has a unique
positive (non-zero) identifier. For the initial
CONNECT request 'id' is zero. After a successful
CONNECT, the server returns in the response header
the identifier to be used for all other requests
(including subsequent CONNECT requests). This is
saved in dev->rmtid.
<p>
'length' specifies the length of the data following the request header.
Currently length is non-zero for READ/WRITE requests.
<p><br>
The server sends an 8 byte response header and maybe some data:
<pre>
+-----+-----+-----+-----+-----+-----+-----+-----+
|code |stat | devnum | id | length |
+-----+-----+-----+-----+-----+-----+-----+-----+
<-------- length --------->
+----- . . . . . -----+
| data |
+----- . . . . . -----+
</pre>
<p>
'code' indicates the response to the request. OK (0x00) indicates
success however other codes also indicate success
but qualified in some manner:
<dl>
<dt>0x80 ERROR<p><dd>
An error occurred. The server provides an error
message in the data section.
<p>
<dt>0x40 IOERR<p><dd>
An i/o error occurred during a READ/WRITE
request. The status byte has the 'unitstat'
data. This should signal the client to issue the
SENSE request to obtain the current sense data.
<p>
<dt>0x20 BUSY<p><dd>
Device was not available for a START request and
the NOWAIT flag bit was turned on.
<p>
<dt>0x10 COMP<p><dd>
Data returned is compressed. The status byte
indicates how the data is compressed (zlib or
bzip2) and at what offset the compressed data
starts (0 .. 15). This bit is only turned on
when both the 'code' and 'status' bytes would
otherwise be zero.
<p>
<dt>0x08 PURGE<p><dd>
START request was issued by the client. A list
of 'records' to be purged from local cache is
returned. These are 'records' that have been
updated since the last START/END request from
the client by other systems. Each record identifier
is a 4-byte field in the data segment. The number
of records then is 'length'/4. If the number of
records exceeds a threshold (16) then 'length'
will be zero indicating that the client should
purge all locally cached records for the device.
</dl>
<p>
'stat' contains status information as a result of the request.
For READ/WRITE requests this contains the 'unitstat'
information if an IOERR occurred.
<p>
'devnum' specifies the server device number
<p>
'id' specifies the system identifier for the request.
<p>
'length' is the size of the data returned.
<p><br>
<h2><a NAME="caching">Caching</a></h2>
<p>
Cached records (eg CKD tracks or FBA blocks) are kept independently on
both the client and server sides. Whenever the client issues a START
request to initiate a channel program the server will return a list
of records to purge from the client's cache that have been updated by
other clients since the last START request. If the list is too large
the server will indicate that the client should purge all records for
the device.
<p><br>
<h2><a NAME="compression">Compression</a></h2>
<p>
Data that would normally be transferred uncompressed between client
and host can optionally be compressed by specifying the '<code>comp=</code>'
keyword on the device configuration statement or attach command.
For example:
<pre class="jcl">
0100 3350 192.168.2.12 comp=3
</pre>
<p>
The value of the 'comp=' keyword is the zlib compression parameter
which should be a number between 1 .. 9. A value closer to 1 means
less compression but less processor time to perform the compression.
A value closer to 9 means the data is compressed more but more processor
time is required.
<p>
If the server is on 'localhost' then you should not specify 'comp='.
Otherwise you are just stealing processor time to do compression/
uncompression from hercules. If the server is on a local network
then I would recommend specifying a low value such as 1, 2 or 3.
We are on a curve here, trying to trade cpu cycles for network traffic
to derive an optimal throughput.
<p>
If the devices on the server are compressed devices (eg CCKD or CFBA)
then the 'records' (eg. track images or block groups) may be transferred
compressed regardless of the 'comp=' setting. This depends on whether
the client supports the compression type (zlib or bzip2) of the record
on the server and whether the record is actually compressed in the
server cache.
<p>
For example:
<p>
Suppose on the client that you execute one or more channel programs
to read a record on a ckd track, update a record on the same track,
and then read another (or the same) record on the track.
<p>
For the first read the server will read the track image and
pass it to the client as it was originally compressed in the file.
To update a portion of the track image the server must uncompress
the track image so data in it can be updated. When the client next
reads from the track image, the track image is uncompressed.
<p>
Specifying 'comp=' means that uncompressed data sent to the client
will be compressed. If the data to be sent to the client is already
compressed then the data is sent as is, unless the client has indicated
that it does not support that compression algorithm.
<p><br>
<h2><a NAME="todo">To Do</a></h2>
<ol>
<li>More doc (sorry, I got winded)
<li>Delays observed during short transfers (redrive select ?)
<li>Better server side behaviour due to disconnect
<li>etc.
</ol>
<p>
Greg Smith
<a href="mailto:gsmith@nc.rr.com"><em>gsmith</em>@<em>nc.rr.com</em></a>
<p><center><hr width=15% noshade>
<a href="index.html"><img src="images/back.gif" border=0 alt="back"></a>
</center>
<p class="lastupd">Last updated $Date$ $Revision$</p>
</BODY>
</HTML>
distrib
>
Fedora
>
17
>
i386
>
media
>
updates
>
by-pkgid
>
8c430bbfe7ce899abe0113d8716ceba1
>
files
>
172
