readme

非常经典的一个分布式系统

*************************************************************************
*                                                                       *
*    Myrinet Express Lustre Networking Driver (MXLND) documentation     *
*                                                                       *
*************************************************************************

README of MXLND

MXLND provides support for Myricom's Myrinet Express (MX) communication
layer in Lustre.

MXLND may be used with either MX-10G or MX-2G. See MX's README for
supported NICs.

Table of Contents:
    I. Installation
       1. Configuring and compiling
       2. Module Parameters
   II. MXLND Performance
  III. Caveats
       1. Systems with different page sizes
       2. Multi-homing
       3. MX endpoint collision
   IV. License
    V. Support

================
I. Installation
================

MXLND is supported on Linux 2.6. It may be possible to run it on 2.4,
but it has not been tested. MXLND requires Myricom's MX version 1.2.1
or higher. See MX's README for the supported list of processors.

1. Configuring and compiling

MXLND should be already integrated into the Lustre build process. To 
build MXLND, you will need to set the path to your MX installation
in Lustre's ./configure:

    --with-mx=/opt/mx

replacing /opt with the actual path. Configure will check to ensure that
the MX version has the required functions. If not, it will fail to build.
To check if MXLND built, look for:

    checking whether to enable Myrinet MX support... yes

in configure's output or the presence of Makefile in
$LUSTRE/lnet/klnds/mxlnd.

2. Module Parameters

MXLND supports a number of load-time parameters using Linux's module
parameter system. On our test systems, we created the following file:

    /etc/modprobe.d/kmxlnd

On some (older?) systems, you may need to modify /etc/modprobe.conf.

The available options are:

    n_waitd     # of completion daemons
    max_peers   maximum number of peers that may connect
    cksum       set non-zero to enable small message (< 4KB) checksums
    ntx         # of total tx message descriptors
    credits     # concurrent sends to a single peer
    board       index value of the Myrinet board (NIC)
    ep_id       MX endpoint ID
    polling     Use 0 to block (wait). A value > 0 will poll that many times before blocking
    hosts       IP-to-hostname resolution file

Of these, only hosts is required. It must be the absolute path to the
MXLND hosts file. For example:

    options kmxlnd hosts=/etc/hosts.mxlnd

The file format for the hosts file is as follows:

IP  HOST  BOARD   EP_ID

The values must be space and/or tab separated where:

    IP is a valid IPv4 address
    HOST is the name returned by `hostname` on that machine
    BOARD is the index of the Myricom NIC (0 for the first card, etc.)
    EP_ID is the MX endpoint ID

You may want to vary the remaining options to obtain the optimal performance
for your platform.

    n_waitd sets the number of threads that process completed MX requests
(sends and receives). In our testing, the default of 1 performed best.

    max_peers tells MXLND the upper limit of machines that it will need to 
communicate with. This affects how many receives it will pre-post and each
receive will use one page of memory. Ideally, on clients, this value will
be equal to the total number of Lustre servers (MDS and OSS). On servers,
it needs to equal the total number of machines in the storage system.

    cksum turns on small message checksums. It can be used to aid in trouble-
shooting. MX also provides an optional checksumming feature which can check 
all messages (large and small). See the MX README for details.

    ntx is the number of total sends in flight from this machine. In actuality,
MXLND reserves half of them for connect messages so make this value twice as large
as you want for the total number of sends in flight.

    credits is the number of in-flight messages for a specific peer. This is part
of the flow-control system in Lustre. Increasing this value may improve performance
but it requires more memory since each message requires at least one page.

    board is the index of the Myricom NIC. Hosts can have multiple Myricom NICs
and this identifies which one MXLND should use. This value must match the board
value in your MXLND hosts file for this host.

    ep_id is the MX endpoint ID. Each process that uses MX is required to have at
least one MX endpoint to access the MX library and NIC. The ID is a simple index
starting at 0. This value must match the endpoint ID value in your MXLND hosts 
file for this host.

    polling determines whether this host will poll or block for MX request com-
pletions. A value of 0 blocks and any positive value will poll that many times
before blocking. Since polling increases CPU usage, we suggest you set this to
0 on the client and experiment with different values for servers.

=====================
II. MXLND Performance
=====================

On MX-2G systems, MXLND should easily saturate the link and use minimal CPU 
(5-10% for read and write operations). On MX-10G systems, MXLND can saturate 
the link and use moderate CPU resources (20-30% for read and write operations).
MX-10G relies on PCI-Express which is relatively new and performance varies
considerably by processor, motherboard and PCI-E chipset. Refer to Myricom's
website for the latest DMA read/write performance results by motherboard. The
DMA results will place an upper-bound on MXLND performance.

============
III. Caveats
============

1. Systems with different page sizes

MXLND will set the maximum small message size equal to the kernel's page size.
This means that machines running MXLND that have different page sizes are not
able to communicate with each other. If you wish to run MXLND in this case,
send email to [e-mail].

2. Multi-homing

At this time, the MXLND cannot drive more than one interface at a time.  Thus,
a single Lustre router cannot route between two MX-10G, between two MX-2G, or
between MX-10G and MX-2G fabrics.

3. MX endpoint collision

Each process that uses MX is required to have at least one MX endpoint to
access the MX library and NIC. Other processes may need to use MX and no two
processes can use the same endpoint ID.  MPICH-MX dynamically chooses one at
MPI startup and should not interfere with MXLND. Sockets-MX, on the other hand,
is hard coded to use 0 for its ID. If it is possible that anyone will want to
run Sockets-MX on this system, use a non-0 value for MXLND's endpoint ID.


===========
IV. License
===========

MXLND is copyright (C) 2006 of Myricom, Inc. 

MXLND is part of Lustre, http://www.lustre.org.

MXLND is free software; you can redistribute it and/or modify it under the
terms of version 2 of the GNU General Public License as published by the Free
Software Foundation.

MXLND is distributed in the hope that it will be useful, but WITHOUT ANY
WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A
PARTICULAR PURPOSE.  See the GNU General Public License for more details.

You should have received a copy of the GNU General Public License along with
Lustre; if not, write to the Free Software Foundation, Inc., 675 Mass Ave,
Cambridge, MA 02139, USA.

==========
V. Support
==========

If you have questions about MXLND, please contact [e-mail].