perftest/SOURCES/ib_atomic_bw.1
2021-10-08 14:29:44 +00:00

299 lines
9.5 KiB
Groff

.\" Copyright (c) 2014, Jan Chaloupka <jchaloup@redhat.com>
.\"
.\" %%%LICENSE_START(GPLv2+_DOC_FULL)
.\" This is free documentation; you can redistribute it and/or
.\" modify it under the terms of the GNU General Public License as
.\" published by the Free Software Foundation; either version 2 of
.\" the License, or (at your option) any later version.
.\"
.\" The GNU General Public License's references to "object code"
.\" and "executables" are to be interpreted as the output of any
.\" document formatting or typesetting system, including
.\" intermediate and printed output.
.\"
.\" This manual 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 this manual; if not, see
.\" <http://www.gnu.org/licenses/>.
.\" %%%LICENSE_END
.TH "IB_ATOMIC_BW" 1 2014 "Open Fabrics Enterprise Distribution"
.\" IB_ATOMIC_BW
.SH NAME
ib_atomic_bw, ib_atomic_lat, ib_read_bw, ib_read_lat, ib_send_bw,
ib_send_lat, ib_write_bw, ib_write_lat
\- Collection of tests written over uverbs intended for use as a
performance micro-benchmark
.SH SYNOPSIS
.sp
.B ib_atomic_bw [<host>] [options]
.sp
.B ib_atomic_lat [<host>] [options]
.sp
.B ib_read_bw [<host>] [options]
.sp
.B ib_read_lat [<host>] [options]
.sp
.B ib_write_bw [<host>] [options]
.sp
.B ib_write_lat [<host>] [options]
.SH DESCRIPTION
This is a collection of tests written over uverbs intended for use as a
performance micro-benchmark. As an example, the tests can be used for
HW or SW tuning and/or functional testing.
The collection conatains a set of BW and latency benchmark such as :
.sp
* Read - ib_read_bw and ib_read_lat.
.sp
* Write - ib_write_bw and ib_wriet_lat.
.sp
* Send - ib_send_bw and ib_send_lat.
.sp
* Atomic - ib_atomic_bw and ib_atomic_lat
.sp
* Raw Etherent (when working with MOFED2) - raw_ethernet_bw, raw_ethernet_lat
The benchmark used the CPU cycle counter to get time stamps without context
switch. Some CPU architectures (e.g., Intel's 80486 or older PPC) do NOT
have such capability.
The latency benchmarks measures round-trip time but reports half of that as
one-way latency.
This means that it may not be sufficiently accurate for asymmetrical
configurations.
On Bw benchmarks, we calculate the BW on send side only, as he calculates
the Bw after collecting completion from the receive side.
In case we use the bidirectional flag , BW is calculated on both sides.
in ib_send_bw, server side also calculate the received throughput.
Min/Median/Max result is reported in latency tests.
The median (vs average) is less sensitive to extreme scores.
Typically, the "Max" value is the first value measured.
Larger samples help marginally only. The default (1000) is pretty good.
Note that an array of cycles_t (typically unsigned long) is allocated
once to collect samples and again to store the difference between them.
Really big sample sizes (e.g., 1 million) might expose other problems
with the program. In this case you can use -N flag (No Peak) to instruct
the test sample only 2 times (begining and end).
All throughput tests now have duration feature as well (-D <seconds to run>)
to instruct the test to run for <seconds to run>.
Another feature added is --run_infinitely, which instruct the test to run
all te time and print throughput every 5 seconds.
The "-H" option (latency) will dump the histogram for additional statistical
analysis.
See xgraph, ygraph, r-base (http://www.r-project.org/), pspp, or other
statistical math programs.
Architectures tested: i686, x86_64, ia64
.SH OPTIONS
The SAME OPTIONS must be passed to both server and client.
If
.I <host>
is not presented, command starts a server and waits for connection.
If it is, command connects to server at
.I <host>.
.sp
.B Common Options:
.RS 4
.TP
\fB\-h\fR, \fB\-\-help\fR
Display this help message screen.
.TP
\fB\-p\fR, \fB\-\-port\fR=\fI<port>\fR
Listen on/connect to port <port> (default: 18515) when exchaning data.
.TP
\fB\-R\fR, \fB\-\-rdma_cm\fR
Connect QPs with rdma_cm and run test on those QPs.
.TP
\fB\-z\fR, \fB\-\-com_rdma_cm\fR
Communicate with rdma_cm module to exchange data \- use regular QPs.
.TP
\fB\-m\fR, \fB\-\-mtu\fR=\fI<mtu>\fR
QP Mtu size (default: active_mtu from ibv_devinfo).
.TP
\fB\-c\fR, \fB\-\-connection\fR=\fI<RC/UC/UD>\fR
Connection type RC/UC/UD (default RC)
.TP
\fB\-d\fR, \fB\-\-ib\-dev\fR=\fI<dev>\fR
Use IB device <dev> (default: first device found).
.TP
\fB\-i\fR, \fB\-\-ib\-port\fR=\fI<port>\fR
Use port <port> of IB device (default: 1).
.TP
\fB\-s\fR, \fB\-\-size\fR=\fI<size>\fR
Size of message to exchange (default: 1).
.TP
\fB\-a\fR, \fB\-\-all\fR
Run sizes from 2 till 2^23.
.TP
\fB\-n\fR, \fB\-\-iters\fR=\fI<iters>\fR
Number of exchanges (at least 100, default: 1000).
.TP
\fB\-x\fR, \fB\-\-gid\-index\fR=\fI<index>\fR
Test uses GID with GID index taken from command
.TP
\fB\-V\fR, \fB\-\-version\fR
Display version number.
.TP
\fB\-e\fR, \fB\-\-events\fR
Sleep on CQ events (default poll).
.TP
\fB\-F\fR, \fB\-\-CPU\-freq\fR
Do not fail even if cpufreq_ondemand module.
.TP
\fB\-I\fR, \fB\-\-inline_size\fR=\fI<size>\fR
Max size of message to be sent in inline mode.
.TP
\fB\-u\fR, \fB\-\-qp\-timeout\fR=\fI<timeout>\fR
QP timeout, timeout value is 4 usec*2 ^timeout (default: 14).
.TP
\fB\-S\fR, \fB\-\-sl\fR=\fI<sl>\fR
SL \- Service Level (default 0)
.TP
\fB\-r\fR, \fB\-\-rx\-depth\fR=\fI<dep>\fR
Make rx queue bigger than tx (default 600).
.RE
.sp
.B Latenct tests options:
.RS 4
.TP
\fB\-C\fR, \fB\-\-report\-cycles\fR
Report times in cpu cycle units.
.TP
\fB\-H\fR, \fB\-\-report\-histogram\fR
Print out all results (Default: summary only).
.TP
\fB\-U\fR, \fB\-\-report\-unsorted\fR
Print out unsorted results (default sorted).
.RE
.sp
.B BW tests options:
.RS 4
.TP
\fB\-b\fR, \fB\-\-bidirectional\fR
Measure bidirectional bandwidth (default uni).
.TP
\fB\-N\fR, \fB\-\-no\fR
peak\-bw Cancel peak\-bw calculation (default with peak\-bw)
.TP
\fB\-Q\fR, \fB\-\-cq\-mod\fR
Generate Cqe only after <cq\-mod> completion
.TP
\fB\-t\fR, \fB\-\-tx\-depth=<dep>\fR
Size of tx queue (default: 128).
.TP
\fB\-O\fR, \fB\-\-dualport\fR
Run test in dual\-port mode (2 QPs). both ports must be active (default OFF).
.TP
\fB\-D\fR, \fB\-\-duration=<sec>\fR
Run test for <sec> period of seconds.
.TP
\fB\-f\fR, \fB\-\-margin=<sec>\fR
When in Duration, measure results within margins (default: 2)
.TP
\fB\-l\fR, \fB\-\-post_list=<list_size>\fR
Post list of WQEs of <list size> size (instead of single post).
.TP
\fB\-q\fR, \fB\-\-qp=<num_of_qps>\fR
Num of QPs running in the process (default: 1).
.TP
\fB\-\-run_infinitely \fR
Run test forever\fR, \fBprint results every 5 seconds.
.RE
.sp
.B SEND tests options:
.RS 4
.TP
\fB\-r\fR, \fB\-\-rx\-depth=<dep>\fR
Size of RX queue (default: 512 in BW test).
.TP
\fB\-g\fR, \fB\-\-mcg=<num_of_qps>\fR
Send messages to multicast group with <num_of_qps> qps attached to it.
.TP
\fB\-M\fR, \fB\-\-MGID=<multicast_gid>\fR
In multicast, uses <multicast_gid> as the group MGID.
.RE
.sp
.B Raw Ethernet BW test options:
.RS 4
.TP
\fB\-A\fR, \fB\-\-atomic_type=<type>\fR
type of atomic operation from {CMP_AND_SWAP,FETCH_AND_ADD}.
.TP
\fB\-o\fR, \fB\-\-outs=<num>\fR
Number of outstanding read/atomic requests \- also on READ tests.
.TP
\fB\-B\fR, \fB\-\-source_mac\fR
source MAC address by this format XX:XX:XX:XX:XX:XX (default take the MAC address form GID).
.TP
\fB\-E\fR, \fB\-\-dest_mac\fR
destination MAC address by this format XX:XX:XX:XX:XX:XX **MUST** be entered.
.TP
\fB\-J\fR, \fB\-\-server_ip\fR
server ip address by this format X.X.X.X (using to send packets with IP header).
.TP
\fB\-j\fR, \fB\-\-client_ip\fR
client ip address by this format X.X.X.X (using to send packets with IP header).
.TP
\fB\-K\fR, \fB\-\-server_port\fR
server udp port number (using to send packets with UPD header).
.TP
\fB\-k\fR, \fB\-\-client_port\fR
client udp port number (using to send packets with UDP header).
.TP
\fB\-Z\fR, \fB\-\-server\fR
choose server side for the current machine (\-\-server/\-\-client must be selected ).
.TP
\fB\-P\fR, \fB\-\-client\fR
choose client side for the current machine (\-\-server/\-\-client must be selected).
.RE
.SH ENVIRONMENT
.B Prerequisites:
.RS
kernel 2.6
.RE
.RS
(kernel module) matches libibverbs
.RE
.RS
(kernel module) matches librdmacm
.RE
.RS
(kernel module) matches libibumad
.RE
.RS
(kernel module) matches libmath (lm).
.RE
.SH NOTES
You need to be running a Subnet Manager on the switch or on one of the nodes in your fabric, in case you are in IB fabric.
.SH BUGS
1. Multicast feauture in ib_send_lat and in ib_send_bw still have many problems!
Will increase the support and bug fixes in this Q, but now the tests may stuck
and could produce undefine behaviours.
.sp
2. Bidirectional feature in ib_send_bw test, when running in UD or UC mode.
The algorithm we use for the bidirectional measurement is designed for RC connection type.
When running in UC or UD connection types, there is a small probablity the test will be stuck.
.sp
3. RDMA_CM feature in read tests still doesn't work.
.sp
4. Dual-port support currently works only with ib_write_bw.
.sp
5. Compabilty issues may occur between different versions of perftest.
Please make sure you work with the same version on both sides to ensure
consistency of the test.
.SH AUTHORS
Please post results/observations to the openib-general mailing list.
See "Contact Us" at http://openib.org/mailman/listinfo/openib-general and
http://www.openib.org.