intel_rdt plugin¶


Plugin name	`intel_rdt`
Type	read
Callbacks	config, init, read, shutdown
Status	supported
Since	5.7
Copyright	2016–2018 Intel Corporation, Serhiy Pshyk , Mateusz Starzyk , Wojciech Andralojc , Michał Aleksiński
Configuration	collectd.conf(5)

The intel_rdt plugin collects information provided by monitoring features of Intel Resource Director Technology (Intel(R) RDT) like Cache Monitoring Technology (CMT), Memory Bandwidth Monitoring (MBM), Cache Allocation Technology (CAT) and Code and Data Prioritization (CDP) Technology provide the hardware framework to monitor and control the utilization of shared resources, like last level cache, memory bandwidth. These Technologies comprise Intel’s Resource Director Technology (RDT). As multithreaded and multicore platform architectures emerge, running workloads in single-threaded, multithreaded, or complex virtual machine environment, the last level cache and memory bandwidth are key resources to manage. Intel introduces CMT, MBM, CAT and CDP to manage these various workloads across shared resources.

Name	Type	Type Instance	Description	Comment
LLC	bytes	llc	last level cache occupancy (CMT)	Existing type
MBL	memory_bandwidth	local	the bandwidth of accessing memory associated with the local socket (MBM)	Existing type
MBR	memory_bandwidth	remote	the bandwidth of accessing the remote socket (MBM)	Existing type
IPC	ipc		instructions per clock	New type introduced in types.db

For a full description of available options please refer to the collectd.conf(5) manual page.

Synopsis¶

 <Plugin "intel_rdt">
   Cores "0-2" "3,4,6" "8-10,15"
 </Plugin>

Parameters¶

Name	Description	Comment
Interval	The interval within which to retrieve statistics on monitored events in seconds	Interval option is supported by collectd and is defined in block. No additional functionality should be developed in intel_rdt plugin to support this option.
Cores	Core groups definition. Monitored metrics are reported as aggregated statistics per group.	The field is represented as list of strings with core group values. Each string represents a list of cores in a group. Allowed formats are: “0,1,2,3” “0-10,20-18” “1,3,5-8,10,0x10-12”. If an empty string is provided as value for this field default cores configuration should be applied - a separate group for each core.

Metrics¶

Metric/Feature/Input	Name	Date Type	Format Example	Description	Dependencies	Limitations	Comments
Metric	Memory Bandwidth on Local NUMA Node	Bytes/Second	3934325	Memory bandwidth utilization by the relevant CPU core on the local NUMA memory channel	PQOS ToolSet	Does not provide the value per process basis due to lack of resctrl fs support	Dependent on PQOS toolset to read the metric value
Metric	Memory Bandwidth on Remote NUMA Node	Bytes/Second	3934325	Memory bandwidth utilization by the relevant CPU core on the remote NUMA memory channel	PQOS ToolSet	Does not provide the value per process basis due to lack of resctrl fs support	Dependent on PQOS toolset to read the metric value
Metric	Total Memory Bandwidth	Bytes/Second	3934325	Total memory bandwidth utilized by a CPU core on local and remote NUMA memory channels	PQOS ToolSet	Does not provide the value per process basis due to lack of resctrl fs support	Not part of the Collectd plugin
Metric	L3 Cache Occupancy	Bytes	45345434	Total Last Level Cache occupancy by a CPU core	PQOS ToolSet	Does not provide the value per process basis due to lack of resctrl fs support	Dependent on PQOS toolset to read the metric value
Metric	Instructions Per Cycle	Integer	23734	Total instructions per cycle executed by a CPU core	PQOS ToolSet	None	Dependent on PQOS toolset to read the metric value
Input	Cores	Integer Array	[0-12] or [1,2,3]	The list of CPU core(s) to be provided as input by the user for which the corresponding metrics are required	None	None	Configuration input in the plugin .conf file
Input	Configuration Interval	Integer	1 or 10	The interval in seconds at which the metrics need to be collectd	None	None	Configuration input in the plugin .conf file
Metric	Memory Bandwidth of the process on Local NUMA Node	Bytes	3934325	Memory bandwidth utilization by the relevant process on the local NUMA memory channel	resctrl fs & kernel 4.14	TBD	Still to create Collectd PR
Metric	Memory Bandwidth of the process on Remote NUMA Node	Bytes	3934325	Memory bandwidth utilization by the relevant process on the remote NUMA memory channel	resctrl fs & kernel 4.14	TBD	Still to create Collectd PR
Metric	Total Memory Bandwidth of the process	Bytes	3934325	Total memory bandwidth utilized by a process on local and remote NUMA memory channels	resctrl fs & kernel 4.14	TBD	Still to create Collectd PR
Metric	L3 Cache Occupancy of the process	Bytes	45345434	Total Last Level Cache occupancy by a process	resctrl fs & kernel 4.14	TBD	Still to create Collectd PR
Metric	Instructions Per Cycle of the process	Integer	23734	Total instructions per cycle executed by a process	resctrl fs & kernel 4.14	TBD	Still to create Collectd PR
Input	Process List	Array	[qemu, pmd]	List of processes for which metrics are to be collected	resctrl fs & kernel 4.14	TBD	Still to create Collectd PR

Example graph¶

Dependencies¶

intel-cmt-cat

Configuration¶

Configuration options · from collectd.conf(5)

See this section in the full configuration reference

The intel_rdt plugin collects information provided by monitoring features of Intel Resource Director Technology (Intel(R) RDT) like Cache Monitoring Technology (CMT), Memory Bandwidth Monitoring (MBM). These features provide information about utilization of shared resources. CMT monitors last level cache occupancy (LLC). MBM supports two types of events reporting local and remote memory bandwidth. Local memory bandwidth (MBL) reports the bandwidth of accessing memory associated with the local socket. Remote memory bandwidth (MBR) reports the bandwidth of accessing the remote socket. Also this technology allows to monitor instructions per clock (IPC). Monitor events are hardware dependant. Monitoring capabilities are detected on plugin initialization and only supported events are monitored.

Note: intel_rdt plugin is using model-specific registers (MSRs), which require an additional capability to be enabled if collectd is run as a service. Please refer to contrib/systemd.collectd.service file for more details.

Synopsis:

<Plugin "intel_rdt">
  Cores "0-2" "3,4,6" "8-10,15"
  Processes "sshd,qemu-system-x86" "bash"
</Plugin>

Options:

Interval seconds

The interval within which to retrieve statistics on monitored events in seconds. For milliseconds divide the time by 1000 for example if the desired interval is 50ms, set interval to 0.05. Due to limited capacity of counters it is not recommended to set interval higher than 1 sec.

Cores cores groups

Monitoring of the events can be configured for group of cores (aggregated statistics). This field defines groups of cores on which to monitor supported events. The field is represented as list of strings with core group values. Each string represents a list of cores in a group. Allowed formats are: 0,1,2,3 0-10,20-18 1,3,5-8,10,0x10-12

If an empty string is provided as value for this field default cores configuration is applied - a separate group is created for each core.

Processes process names groups

Monitoring of the events can be configured for group of processes (aggregated statistics). This field defines groups of processes on which to monitor supported events. The field is represented as list of strings with process names group values. Each string represents a list of processes in a group. Allowed format is: sshd,bash,qemu

Note: By default global interval is used to retrieve statistics on monitored events. To configure a plugin specific interval use Interval option of the intel_rdt <LoadPlugin> block. For milliseconds divide the time by 1000 for example if the desired interval is 50ms, set interval to 0.05. Due to limited capacity of counters it is not recommended to set interval higher than 1 sec.