1PQOS(8) System Manager's Manual PQOS(8)
2
3
4
6 pqos, pqos-msr, pqos-os - Intel(R) Resource Director Technology moni‐
7 toring and control tool
8
10 pqos [OPTIONS]...
11
13 Intel(R) Resource Director Technology is designed to monitor and manage
14 cpu resources and improve performance of applications and virtual
15 machines.
16
17 Intel(R) Resource Director Technology includes monitoring and control
18 technologies. Monitoring technologies include CMT (Cache Monitoring
19 Technology), which monitors occupancy of last level cache, and MBM
20 (Memory Bandwidth Monitoring). Control technologies include CAT (Cache
21 Allocation Technology) and CDP (Code Data Prioritization).
22
23 pqos supports CMT and MBM on a per core or hardware thread basis. MBM
24 supports two types of events reporting local and remote memory band‐
25 width.
26 pqos-msr and pqos-os are simple pqos wrapper scripts that automatically
27 select the MSR or OS/Kernel library interface to program the technolo‐
28 gies.
29 Please see the -I option below for more information.
30
31 For hardware information please refer to the README located on:
32 https://github.com/01org/intel-cmt-cat/blob/master/README
33
35 pqos options are as follow:
36
37 -h, --help
38 show help
39
40 -v, --verbose
41 verbose mode
42
43 -V, --super-verbose
44 super-verbose mode
45
46 -l FILE, --log-file=FILE
47 log messages into selected log FILE
48
49 -s, --show
50 show the current allocation and monitoring configuration
51
52 -d, --display
53 display supported Intel(R) Resource Director Technology capabil‐
54 ities
55
56 -D, --display-verbose
57 display supported Intel(R) Resource Director Technology capabil‐
58 ities in verbose mode
59
60 -f FILE, --config-file=FILE
61 load commands from selected configuration FILE
62
63 -e CLASSDEF, --alloc-class=CLASSDEF
64 define the allocation classes on all CPU sockets. CLASSDEF for‐
65 mat is "TYPE:ID=DEFINITION;...".
66 define classes for selected CPU resources. CLASSDEF format is
67 "TYPE[@RESOURCE_ID]:ID=DEFINITION;...".
68 For CAT, TYPE is "llc" for the last level cache (aka l3) and
69 "l2" for level 2 cache, ID is a CLOS number and DEFINITION is a
70 bitmask.
71 For MBA, TYPE is "mba", ID is a CLOS number and DEFINITION is a
72 value between 1 and 100 representing the percentage of available
73 memory bandwidth.
74 RESOURCE_ID is a unique number that can represent a socket or
75 l2/l3 cache identifier. The RESOURCE_ID for each logical CPU can
76 be found using "pqos -s"
77 Note: When L2/L3 CDP is on, ID can be postfixed with 'D' for
78 data or 'C' for code.
79 Note: L2/L3 CDP is available on selected CPUs only.
80 Some examples:
81 "-e
82 llc:0=0xffff;llc:1=0x00ff;llc@0-1:2=0xff00;l2:2=0x3f;l2@2:1=0xf"
83 "-e llc:0d=0xfff;llc:0c=0xfff00"
84 "-e l2:0d=0xf;l2:0c=0xc"
85 "-e mba:1=30;mba@1:3=80"
86 Note:
87 "-e l2:2=0x3f" means that COS2 for all L2 cache clusters
88 is changed to 0x3f.
89 "-e l2@2:1=0xf" means that COS1 for L2 cache cluster 2 is
90 changed to 0xf.
91 "-e mba:1=30" means that COS1, on all sockets, can uti‐
92 lize up to 30% of available memory bandwidth.
93
94 -a CLASS2ID, --alloc-assoc=CLASS2ID
95 associate allocation classes with cores. CLASS2ID format is
96 "TYPE:ID=CORE_LIST;..." or "TYPE:ID=TASK_LIST;...".
97 For CAT, TYPE is "llc", "core" or "pid" and ID is a class num‐
98 ber. CORE_LIST is comma or dash separated list of cores.
99 TASK_LIST is comma or dash separated list of process/task ID's.
100 For example:
101 "-a llc:0=0,2,4,6-10;llc:1=1;" associates cores 0, 2, 4,
102 6, 7, 8, 9, 10 with CAT class 0 and core 1 with class 1.
103 "-a core:0=0,2,4,6-10;core:1=1;" associates cores 0, 2,
104 4, 6, 7, 8, 9, 10 with CAT class 0 and core 1 with class
105 1.
106 "-I -a pid:0=3543,7643,4556;pid:1=7644;" associates
107 process ID 3543, 7643, 4556 with CAT class 0 and process
108 ID 7644 with class 1.
109 Note:
110 The -I option must be used for PID association.
111
112 -R [CONFIG[,CONFIG]], --alloc-reset[=CONFIG[,CONFIG]]
113 reset allocation setting (L3 CAT, L2 CAT) and reconfigure L3
114 CDP. CONFIG is one of the following options:
115 l3cdp-on sets L3 CDP on
116 l3cdp-off sets L3 CDP off
117 l3cdp-any keep current L3 CDP setting (default)
118 l2cdp-on sets L2 CDP on
119 l2cdp-off sets L2 CDP off
120 l2cdp-any keep current L2 CDP setting (default)
121
122 -m EVTCORES, --mon-core=EVTCORES
123 select the cores and events for monitoring, EVTCORES format is
124 "EVENT:CORE_LIST". Valid EVENT settings are:
125 - "llc" for CMT (LLC occupancy)
126 - "mbr" for MBR (remote memory bandwidth)
127 - "mbl" for MBL (local memory bandwidth)
128 - "all" or "" for all detected event types
129 CORE_LIST is comma or dash separated list of cores.
130 Example "-m all:0,2,4-10;llc:1,3;mbr:11-12".
131 Core statistics can be grouped by enclosing the core list in
132 square brackets.
133 Example "-m llc:[0-3];all:[4,5,6];mbr:[0-3],7,8".
134
135 -p [EVTPIDS], --mon-pid[=EVTPIDS]
136 select top 10 most active (CPU utilizing) process ids to monitor
137 or select the process ids and events to monitor, EVTPIDS format
138 is "EVENT:PID_LIST".
139 See -m option for valid EVENT settings. PID_LIST is comma sepa‐
140 rated list of process ids.
141 Examples:
142 "-p llc:22,25673"
143 "-p all:892,4588-4592"
144
145 Process' IDs can be grouped by enclosing them in square brack‐
146 ets.
147 Examples:
148 "-p llc:[22,25673]"
149 "-p all:892,[4588-4592]"
150
151 Note:
152 Requires Linux and kernel versions 4.10 and newer.
153 The -I option must be used for PID monitoring.
154 It is not possible to track both processes and cores at
155 the same time.
156
157 -T, --mon-top
158 enable top like monitoring output sorted by highest LLC occu‐
159 pancy
160
161 -o FILE, --mon-file FILE
162 select output FILE to store monitored data in, the default is
163 'stdout'
164
165 -u TYPE, --mon-file-type=TYPE
166 select the output format TYPE for monitored data. Supported TYPE
167 settings are: "text" (default), "xml" and "csv".
168
169 -i INTERVAL, --mon-interval=INTERVAL
170 define monitoring sampling INTERVAL in 100ms units, 1=100ms,
171 default 10=10x100ms=1s
172
173 -t SECONDS, --mon-time=SECONDS
174 define monitoring time in seconds, use 'inf' or 'infinite' for
175 infinite monitoring. Use CTRL+C to stop monitoring at any time.
176
177 -r, --mon-reset
178 reset monitoring and use all RMID's and cores in the system
179
180 -H, --profile-list
181 list supported allocation profiles
182
183 -c PROFILE, --profile-set=PROFILE
184 select a PROFILE from predefined allocation classes, use -H to
185 list available profiles
186
187 -I, --iface-os
188 set the library interface to use the kernel implementation. If
189 not set the default implementation is to program the MSR's
190 directly.
191
193 CMT, MBM and CAT are configured using Model Specific Registers (MSRs).
194 The pqos software executes in user space, and access to the MSRs is
195 obtained through a standard Linux* interface. The msr file interface is
196 protected and requires root privileges. The msr driver might not be
197 auto-loaded and on some modular kernels the driver may need to be
198 loaded manually:
199
200 For Linux:
201 sudo modprobe msr
202
203 For FreeBSD:
204 sudo kldload cpuctl
205
206 Interface enforcement:
207 If you require system wide interface enforcement you can do so by set‐
208 ting the "RDT_IFACE" environment variable.
209
211 msr(4)
212
214 pqos was written by Tomasz Kantecki <tomasz.kantecki@intel.com>, Marcel
215 Cornu <marcel.d.cornu@intel.com>, Aaron Hetherington <aaron.hethering‐
216 ton@intel.com>
217
218 This is free software; see the source for copying conditions. There is
219 NO warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR
220 PURPOSE.
221
222
223
224 November 16, 2017 PQOS(8)