With Sub-NUMA Cluster mode enabled the scope of monitoring resources is
per-NODE instead of per-L3 cache. Suffixes of directories with "L3" in
their name refer to Sub-NUMA nodes instead of L3 cache ids.
Users should be aware that SNC mode also affects the amount of L3 cache
available for allocation within each SNC node.
Signed-off-by: Tony Luck <[email protected]>
---
Changes since v5:
Added addtional details about challenges tracking tasks when SNC
mode is enabled.
---
Documentation/arch/x86/resctrl.rst | 34 +++++++++++++++++++++++++++---
1 file changed, 31 insertions(+), 3 deletions(-)
diff --git a/Documentation/arch/x86/resctrl.rst b/Documentation/arch/x86/resctrl.rst
index cb05d90111b4..d6b6a4cfd967 100644
--- a/Documentation/arch/x86/resctrl.rst
+++ b/Documentation/arch/x86/resctrl.rst
@@ -345,9 +345,15 @@ When control is enabled all CTRL_MON groups will also contain:
When monitoring is enabled all MON groups will also contain:
"mon_data":
- This contains a set of files organized by L3 domain and by
- RDT event. E.g. on a system with two L3 domains there will
- be subdirectories "mon_L3_00" and "mon_L3_01". Each of these
+ This contains a set of files organized by L3 domain or by NUMA
+ node (depending on whether Sub-NUMA Cluster (SNC) mode is disabled
+ or enabled respectively) and by RDT event. E.g. on a system with
+ SNC mode disabled with two L3 domains there will be subdirectories
+ "mon_L3_00" and "mon_L3_01". The numerical suffix refers to the
+ L3 cache id. With SNC enabled the directory names are the same,
+ but the numerical suffix refers to the node id.
+ Mappings from node ids to CPUs are available in the
+ /sys/devices/system/node/node*/cpulist files. Each of these
directories have one file per event (e.g. "llc_occupancy",
"mbm_total_bytes", and "mbm_local_bytes"). In a MON group these
files provide a read out of the current value of the event for
@@ -452,6 +458,28 @@ and 0xA are not. On a system with a 20-bit mask each bit represents 5%
of the capacity of the cache. You could partition the cache into four
equal parts with masks: 0x1f, 0x3e0, 0x7c00, 0xf8000.
+Notes on Sub-NUMA Cluster mode
+==============================
+When SNC mode is enabled the "llc_occupancy", "mbm_total_bytes", and
+"mbm_local_bytes" will only give meaningful results for well behaved NUMA
+applications. I.e. those that perform the majority of memory accesses
+to memory on the local NUMA node to the CPU where the task is executing.
+Note that Linux may load balance tasks between Sub-NUMA nodes much
+more readily than between regular NUMA nodes since the CPUs on SNC
+share the same L3 cache and the system may report the NUMA distance
+between SNC nodes with a lower value than used for regular NUMA nodes.
+Tasks that migrate between nodes will have their traffic recorded by the
+counters in different SNC nodes so a user will need to read mon_data
+files from each node on which the task executed to get the full
+view of traffic for which the task was the source.
+
+
+The cache allocation feature still provides the same number of
+bits in a mask to control allocation into the L3 cache. But each
+of those ways has its capacity reduced because the cache is divided
+between the SNC nodes. The values reported in the resctrl
+"size" files are adjusted accordingly.
+
Memory bandwidth Allocation and monitoring
==========================================
--
2.41.0
Hi Tony,
On Thu, Sep 28, 2023 at 9:14 PM Tony Luck <[email protected]> wrote:
> diff --git a/Documentation/arch/x86/resctrl.rst b/Documentation/arch/x86/resctrl.rst
> index cb05d90111b4..d6b6a4cfd967 100644
> --- a/Documentation/arch/x86/resctrl.rst
> +++ b/Documentation/arch/x86/resctrl.rst
> @@ -345,9 +345,15 @@ When control is enabled all CTRL_MON groups will also contain:
> When monitoring is enabled all MON groups will also contain:
>
> "mon_data":
> - This contains a set of files organized by L3 domain and by
> - RDT event. E.g. on a system with two L3 domains there will
> - be subdirectories "mon_L3_00" and "mon_L3_01". Each of these
> + This contains a set of files organized by L3 domain or by NUMA
> + node (depending on whether Sub-NUMA Cluster (SNC) mode is disabled
> + or enabled respectively) and by RDT event. E.g. on a system with
> + SNC mode disabled with two L3 domains there will be subdirectories
> + "mon_L3_00" and "mon_L3_01". The numerical suffix refers to the
> + L3 cache id. With SNC enabled the directory names are the same,
> + but the numerical suffix refers to the node id.
> + Mappings from node ids to CPUs are available in the
> + /sys/devices/system/node/node*/cpulist files. Each of these
The explanation of mon_data seems overwhelmingly SNC-centric now.
Maybe the SNC section should be responsible for explaining its impact
on the mon_data directory. Mainly by reminding the reader that domain
ids in the mon_data directory are node ids in SNC mode.
> directories have one file per event (e.g. "llc_occupancy",
> "mbm_total_bytes", and "mbm_local_bytes"). In a MON group these
> files provide a read out of the current value of the event for
> @@ -452,6 +458,28 @@ and 0xA are not. On a system with a 20-bit mask each bit represents 5%
> of the capacity of the cache. You could partition the cache into four
> equal parts with masks: 0x1f, 0x3e0, 0x7c00, 0xf8000.
>
> +Notes on Sub-NUMA Cluster mode
> +==============================
> +When SNC mode is enabled the "llc_occupancy", "mbm_total_bytes", and
> +"mbm_local_bytes" will only give meaningful results for well behaved NUMA
> +applications. I.e. those that perform the majority of memory accesses
> +to memory on the local NUMA node to the CPU where the task is executing.
Not being specific about why the results aren't meaningful, this
sounds vague and alarming.
> +Note that Linux may load balance tasks between Sub-NUMA nodes much
> +more readily than between regular NUMA nodes since the CPUs on SNC
> +share the same L3 cache and the system may report the NUMA distance
> +between SNC nodes with a lower value than used for regular NUMA nodes.
> +Tasks that migrate between nodes will have their traffic recorded by the
> +counters in different SNC nodes so a user will need to read mon_data
> +files from each node on which the task executed to get the full
> +view of traffic for which the task was the source.
> +
> +
> +The cache allocation feature still provides the same number of
> +bits in a mask to control allocation into the L3 cache. But each
> +of those ways has its capacity reduced because the cache is divided
> +between the SNC nodes. The values reported in the resctrl
> +"size" files are adjusted accordingly.
> +
> Memory bandwidth Allocation and monitoring
> ==========================================
>
> --
> 2.41.0
>
Reviewed-by: Peter Newman <[email protected]>
On Fri, Sep 29, 2023 at 04:54:21PM +0200, Peter Newman wrote:
> Hi Tony,
>
> On Thu, Sep 28, 2023 at 9:14 PM Tony Luck <[email protected]> wrote:
> > diff --git a/Documentation/arch/x86/resctrl.rst b/Documentation/arch/x86/resctrl.rst
> > index cb05d90111b4..d6b6a4cfd967 100644
> > --- a/Documentation/arch/x86/resctrl.rst
> > +++ b/Documentation/arch/x86/resctrl.rst
> > @@ -345,9 +345,15 @@ When control is enabled all CTRL_MON groups will also contain:
> > When monitoring is enabled all MON groups will also contain:
> >
> > "mon_data":
> > - This contains a set of files organized by L3 domain and by
> > - RDT event. E.g. on a system with two L3 domains there will
> > - be subdirectories "mon_L3_00" and "mon_L3_01". Each of these
> > + This contains a set of files organized by L3 domain or by NUMA
> > + node (depending on whether Sub-NUMA Cluster (SNC) mode is disabled
> > + or enabled respectively) and by RDT event. E.g. on a system with
> > + SNC mode disabled with two L3 domains there will be subdirectories
> > + "mon_L3_00" and "mon_L3_01". The numerical suffix refers to the
> > + L3 cache id. With SNC enabled the directory names are the same,
> > + but the numerical suffix refers to the node id.
> > + Mappings from node ids to CPUs are available in the
> > + /sys/devices/system/node/node*/cpulist files. Each of these
>
> The explanation of mon_data seems overwhelmingly SNC-centric now.
> Maybe the SNC section should be responsible for explaining its impact
> on the mon_data directory. Mainly by reminding the reader that domain
> ids in the mon_data directory are node ids in SNC mode.
I cut out all the examples and just note that the numerical suffices
are nodes instead of cache instances.
This bit of the git diff now reads:
- This contains a set of files organized by L3 domain and by
- RDT event. E.g. on a system with two L3 domains there will
- be subdirectories "mon_L3_00" and "mon_L3_01". Each of these
+ This contains a set of files organized by L3 domain or by NUMA
+ node (depending on whether Sub-NUMA Cluster (SNC) mode is disabled
+ or enabled respectively) and by RDT event. Each of these
>
>
> > directories have one file per event (e.g. "llc_occupancy",
> > "mbm_total_bytes", and "mbm_local_bytes"). In a MON group these
> > files provide a read out of the current value of the event for
> > @@ -452,6 +458,28 @@ and 0xA are not. On a system with a 20-bit mask each bit represents 5%
> > of the capacity of the cache. You could partition the cache into four
> > equal parts with masks: 0x1f, 0x3e0, 0x7c00, 0xf8000.
> >
> > +Notes on Sub-NUMA Cluster mode
> > +==============================
> > +When SNC mode is enabled the "llc_occupancy", "mbm_total_bytes", and
> > +"mbm_local_bytes" will only give meaningful results for well behaved NUMA
> > +applications. I.e. those that perform the majority of memory accesses
> > +to memory on the local NUMA node to the CPU where the task is executing.
>
> Not being specific about why the results aren't meaningful, this
> sounds vague and alarming.
Removed the trigger word "meaningful" and re-worded to just explain
the increased liklihood that tasks will migrate between nodes, so users
must collect data from all nodes. Technically this has always been true
on multi-socket systems. But since there is a much higher barrier to
task migration between sockets, users may find that simple measurements
that used to work now behave differently.
New version:
+Notes on Sub-NUMA Cluster mode
+==============================
+When SNC mode is enabled Linux may load balance tasks between Sub-NUMA
+nodes much more readily than between regular NUMA nodes since the CPUs
+on Sub-NUMA nodes share the same L3 cache and the system may report
+the NUMA distance between Sub-NUMA nodes with a lower value than used
+for regular NUMA nodes. Users who do not bind tasks to the CPUs of a
+specific Sub-NUMA node must read the "llc_occupancy", "mbm_total_bytes",
+and "mbm_local_bytes" for all Sub-NUMA nodes where the tasks may execute
+to get the full view of traffic for which the tasks were the source.
+
+The cache allocation feature still provides the same number of
+bits in a mask to control allocation into the L3 cache. But each
+of those ways has its capacity reduced because the cache is divided
+between the SNC nodes. The values reported in the resctrl
+"size" files are adjusted accordingly.
>
> > +Note that Linux may load balance tasks between Sub-NUMA nodes much
> > +more readily than between regular NUMA nodes since the CPUs on SNC
> > +share the same L3 cache and the system may report the NUMA distance
> > +between SNC nodes with a lower value than used for regular NUMA nodes.
> > +Tasks that migrate between nodes will have their traffic recorded by the
> > +counters in different SNC nodes so a user will need to read mon_data
> > +files from each node on which the task executed to get the full
> > +view of traffic for which the task was the source.
> > +
> > +
> > +The cache allocation feature still provides the same number of
> > +bits in a mask to control allocation into the L3 cache. But each
> > +of those ways has its capacity reduced because the cache is divided
> > +between the SNC nodes. The values reported in the resctrl
> > +"size" files are adjusted accordingly.
> > +
> > Memory bandwidth Allocation and monitoring
> > ==========================================
> >
> > --
> > 2.41.0
> >
>
> Reviewed-by: Peter Newman <[email protected]>