github/kubernetes
1
0
Fork 0
mirror of https://github.com/k3s-io/kubernetes.git synced 2025-07-26 13:07:07 +00:00
Code Issues Projects Releases Wiki Activity

Update TopologyManager algorithm for selecting "best" non-preferred hint

Browse Source 
For the 'single-numa' and 'restricted' TopologyManager policies, pods are only
admitted if all of their containers have perfect alignment across the set of
resources they are requesting. The best-effort policy, on the other hand, will
prefer allocations that have perfect alignment, but fall back to a non-preferred
alignment if perfect alignment can't be achieved.

The existing algorithm of how to choose the best hint from the set of
"non-preferred" hints is fairly naive and often results in choosing a
sub-optimal hint. It works fine in cases where all resources would end up
coming from a single NUMA node (even if its not the same NUMA nodes), but
breaks down as soon as multiple NUMA nodes are required for the "best"
alignment.  We will never be able to achieve perfect alignment with these
non-preferred hints, but we should try and do something more intelligent than
simply choosing the hint with the narrowest mask.

In an ideal world, we would have the TopologyManager return a set of
"resources-relative" hints (as opposed to a common hint for all resources as is
done today). Each resource-relative hint would indicate how many other
resources could be aligned to it on a given NUMA node, and a  hint provider
would use this information to allocate its resources in the most aligned way
possible. There are likely some edge cases to consider here, but such an
algorithm would allow us to do partial-perfect-alignment of "some" resources,
even if all resources could not be perfectly aligned.

Unfortunately, supporting something like this would require a major redesign to
how the TopologyManager interacts with its hint providers (as well as how those
hint providers make decisions based on the hints they get back).

That said, we can still do better than the naive algorithm we have today, and
this patch provides a mechanism to do so.

We start by looking at the set of hints passed into the TopologyManager for
each resource and generate a list of the minimum number of NUMA nodes required
to satisfy an allocation for a given resource. Each entry in this list then
contains the 'minNUMAAffinity.Count()' for a given resources. Once we have this
list, we find the *maximum* 'minNUMAAffinity.Count()' from the list and mark
that as the 'bestNonPreferredAffinityCount' that we would like to have
associated with whatever "bestHint" we ultimately generate. The intuition being
that we would like to (at the very least) get alignment for those resources
that *require* multiple NUMA nodes to satisfy their allocation. If we can't
quite get there, then we should try to come as close to it as possible.

Once we have this 'bestNonPreferredAffinityCount', the algorithm proceeds as
follows:

If the mergedHint and bestHint are both non-preferred, then try and find a hint
whose affinity count is as close to (but not higher than) the
bestNonPreferredAffinityCount as possible. To do this we need to consider the
following cases and react accordingly:

  1. bestHint.NUMANodeAffinity.Count() >  bestNonPreferredAffinityCount
  2. bestHint.NUMANodeAffinity.Count() == bestNonPreferredAffinityCount
  3. bestHint.NUMANodeAffinity.Count() <  bestNonPreferredAffinityCount

For case (1), the current bestHint is larger than the
bestNonPreferredAffinityCount, so updating to any narrower mergeHint is
preferred over staying where we are.

For case (2), the current bestHint is equal to the
bestNonPreferredAffinityCount, so we would like to stick with what we have
*unless* the current mergedHint is also equal to bestNonPreferredAffinityCount
and it is narrower.

For case (3), the current bestHint is less than bestNonPreferredAffinityCount,
so we would like to creep back up to bestNonPreferredAffinityCount as close as
we can. There are three cases to consider here:

  3a. mergedHint.NUMANodeAffinity.Count() >  bestNonPreferredAffinityCount
  3b. mergedHint.NUMANodeAffinity.Count() == bestNonPreferredAffinityCount
  3c. mergedHint.NUMANodeAffinity.Count() <  bestNonPreferredAffinityCount

For case (3a), we just want to stick with the current bestHint because choosing
a new hint that is greater than bestNonPreferredAffinityCount would be
counter-productive.

For case (3b), we want to immediately update bestHint to the current
mergedHint, making it now equal to bestNonPreferredAffinityCount.

For case (3c), we know that *both* the current bestHint and the current
mergedHint are less than bestNonPreferredAffinityCount, so we want to choose
one that brings us back up as close to bestNonPreferredAffinityCount as
possible. There are three cases to consider here:

  3ca. mergedHint.NUMANodeAffinity.Count() >  bestHint.NUMANodeAffinity.Count()
  3cb. mergedHint.NUMANodeAffinity.Count() <  bestHint.NUMANodeAffinity.Count()
  3cc. mergedHint.NUMANodeAffinity.Count() == bestHint.NUMANodeAffinity.Count()

For case (3ca), we want to immediately update bestHint to mergedHint because
that will bring us closer to the (higher) value of
bestNonPreferredAffinityCount.

For case (3cb), we want to stick with the current bestHint because choosing the
current mergedHint would strictly move us further away from the
bestNonPreferredAffinityCount.

Finally, for case (3cc), we know that the current bestHint and the current
mergedHint are equal, so we simply choose the narrower of the 2.

This patch implements this algorithm for the case where we must choose from a
set of non-preferred hints and provides a set of unit-tests to verify its
correctness.

Signed-off-by: Kevin Klues <kklues@nvidia.com>
This commit is contained in:
Kevin Klues 2022-02-16 08:18:39 +00:00
parent f8601cb5a3
commit 99c57828ce
2 changed files with 357 additions and 7 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

150
pkg/kubelet/cm/topologymanager/policy.go
View File

@ -94,7 +94,40 @@ func filterProvidersHints(providersHints []map[string][]TopologyHint) [][]Topolo
return allProviderHints
}
func compareHints(current *TopologyHint, candidate *TopologyHint) *TopologyHint {
func narrowestHint(hints []TopologyHint) *TopologyHint {
if len(hints) == 0 {
return nil
}
var narrowestHint *TopologyHint
for i := range hints {
if hints[i].NUMANodeAffinity == nil {
continue
}
if narrowestHint == nil {
narrowestHint = &hints[i]
}
if hints[i].NUMANodeAffinity.IsNarrowerThan(narrowestHint.NUMANodeAffinity) {
narrowestHint = &hints[i]
}
}
return narrowestHint
}
func maxOfMinAffinityCounts(filteredHints [][]TopologyHint) int {
maxOfMinCount := 0
for _, resourceHints := range filteredHints {
narrowestHint := narrowestHint(resourceHints)
if narrowestHint == nil {
continue
}
if narrowestHint.NUMANodeAffinity.Count() > maxOfMinCount {
maxOfMinCount = narrowestHint.NUMANodeAffinity.Count()
}
}
return maxOfMinCount
}
func compareHints(bestNonPreferredAffinityCount int, current *TopologyHint, candidate *TopologyHint) *TopologyHint {
// Only consider candidates that result in a NUMANodeAffinity > 0 to
// replace the current bestHint.
if candidate.NUMANodeAffinity.Count() == 0 {
@ -119,18 +152,121 @@ func compareHints(current *TopologyHint, candidate *TopologyHint) *TopologyHint
return current
}
// If the current bestHint and the candidate hint have the same
// preference, only consider candidate hints that have a narrower
// If the current bestHint and the candidate hint are both preferred,
// then only consider candidate hints that have a narrower
// NUMANodeAffinity than the NUMANodeAffinity in the current bestHint.
if !candidate.NUMANodeAffinity.IsNarrowerThan(current.NUMANodeAffinity) {
if current.Preferred && candidate.Preferred {
if candidate.NUMANodeAffinity.IsNarrowerThan(current.NUMANodeAffinity) {
return candidate
}
return current
}
// In all other cases, update the bestHint to the candidate hint.
return candidate
// The only case left is if the current best bestHint and the candidate
// hint are both non-preferred. In this case, try and find a hint whose
// affinity count is as close to (but not higher than) the
// bestNonPreferredAffinityCount as possible. To do this we need to
// consider the following cases and react accordingly:
//
// 1. current.NUMANodeAffinity.Count() > bestNonPreferredAffinityCount
// 2. current.NUMANodeAffinity.Count() == bestNonPreferredAffinityCount
// 3. current.NUMANodeAffinity.Count() < bestNonPreferredAffinityCount
//
// For case (1), the current bestHint is larger than the
// bestNonPreferredAffinityCount, so updating to any narrower mergeHint
// is preferred over staying where we are.
//
// For case (2), the current bestHint is equal to the
// bestNonPreferredAffinityCount, so we would like to stick with what
// we have *unless* the candidate hint is also equal to
// bestNonPreferredAffinityCount and it is narrower.
//
// For case (3), the current bestHint is less than
// bestNonPreferredAffinityCount, so we would like to creep back up to
// bestNonPreferredAffinityCount as close as we can. There are three
// cases to consider here:
//
// 3a. candidate.NUMANodeAffinity.Count() > bestNonPreferredAffinityCount
// 3b. candidate.NUMANodeAffinity.Count() == bestNonPreferredAffinityCount
// 3c. candidate.NUMANodeAffinity.Count() < bestNonPreferredAffinityCount
//
// For case (3a), we just want to stick with the current bestHint
// because choosing a new hint that is greater than
// bestNonPreferredAffinityCount would be counter-productive.
//
// For case (3b), we want to immediately update bestHint to the
// candidate hint, making it now equal to bestNonPreferredAffinityCount.
//
// For case (3c), we know that *both* the current bestHint and the
// candidate hint are less than bestNonPreferredAffinityCount, so we
// want to choose one that brings us back up as close to
// bestNonPreferredAffinityCount as possible. There are three cases to
// consider here:
//
// 3ca. candidate.NUMANodeAffinity.Count() > current.NUMANodeAffinity.Count()
// 3cb. candidate.NUMANodeAffinity.Count() < current.NUMANodeAffinity.Count()
// 3cc. candidate.NUMANodeAffinity.Count() == current.NUMANodeAffinity.Count()
//
// For case (3ca), we want to immediately update bestHint to the
// candidate hint because that will bring us closer to the (higher)
// value of bestNonPreferredAffinityCount.
//
// For case (3cb), we want to stick with the current bestHint because
// choosing the candidate hint would strictly move us further away from
// the bestNonPreferredAffinityCount.
//
// Finally, for case (3cc), we know that the current bestHint and the
// candidate hint are equal, so we simply choose the narrower of the 2.
// Case 1
if current.NUMANodeAffinity.Count() > bestNonPreferredAffinityCount {
if candidate.NUMANodeAffinity.IsNarrowerThan(current.NUMANodeAffinity) {
return candidate
}
return current
}
// Case 2
if current.NUMANodeAffinity.Count() == bestNonPreferredAffinityCount {
if candidate.NUMANodeAffinity.Count() != bestNonPreferredAffinityCount {
return current
}
if candidate.NUMANodeAffinity.IsNarrowerThan(current.NUMANodeAffinity) {
return candidate
}
return current
}
// Case 3a
if candidate.NUMANodeAffinity.Count() > bestNonPreferredAffinityCount {
return current
}
// Case 3b
if candidate.NUMANodeAffinity.Count() == bestNonPreferredAffinityCount {
return candidate
}
// Case 3ca
if candidate.NUMANodeAffinity.Count() > current.NUMANodeAffinity.Count() {
return candidate
}
// Case 3cb
if candidate.NUMANodeAffinity.Count() < current.NUMANodeAffinity.Count() {
return current
}
// Case 3cc
if candidate.NUMANodeAffinity.IsNarrowerThan(current.NUMANodeAffinity) {
return candidate
}
return current
}
func mergeFilteredHints(numaNodes []int, filteredHints [][]TopologyHint) TopologyHint {
// Set bestNonPreferredAffinityCount to help decide which affinity mask is
// preferred amongst all non-preferred hints. We calculate this value as
// the maximum of the minimum affinity counts supplied for any given hint
// provider. In other words, prefer a hint that has an affinity mask that
// includes all of the NUMA nodes from the provider that requires the most
// NUMA nodes to satisfy its allocation.
bestNonPreferredAffinityCount := maxOfMinAffinityCounts(filteredHints)
var bestHint *TopologyHint
iterateAllProviderTopologyHints(filteredHints, func(permutation []TopologyHint) {
// Get the NUMANodeAffinity from each hint in the permutation and see if any
@ -139,7 +275,7 @@ func mergeFilteredHints(numaNodes []int, filteredHints [][]TopologyHint) Topolog
// Compare the current bestHint with the candidate mergedHint and
// update bestHint if appropriate.
bestHint = compareHints(bestHint, &mergedHint)
bestHint = compareHints(bestNonPreferredAffinityCount, bestHint, &mergedHint)
})
if bestHint == nil {

214
pkg/kubelet/cm/topologymanager/policy_test.go
View File

@ -615,6 +615,220 @@ func (p *bestEffortPolicy) mergeTestCases(numaNodes []int) []policyMergeTestCase
Preferred: false,
},
},
{
name: "bestNonPreferredAffinityCount (1)",
hp: []HintProvider{
&mockHintProvider{
map[string][]TopologyHint{
"resource1": {
{
NUMANodeAffinity: NewTestBitMask(0, 1, 2, 3),
Preferred: false,
},
{
NUMANodeAffinity: NewTestBitMask(0, 1),
Preferred: false,
},
},
},
},
&mockHintProvider{
map[string][]TopologyHint{
"resource2": {
{
NUMANodeAffinity: NewTestBitMask(0, 1),
Preferred: false,
},
},
},
},
},
expected: TopologyHint{
NUMANodeAffinity: NewTestBitMask(0, 1),
Preferred: false,
},
},
{
name: "bestNonPreferredAffinityCount (2)",
hp: []HintProvider{
&mockHintProvider{
map[string][]TopologyHint{
"resource1": {
{
NUMANodeAffinity: NewTestBitMask(0, 1, 2, 3),
Preferred: false,
},
{
NUMANodeAffinity: NewTestBitMask(0, 1),
Preferred: false,
},
},
},
},
&mockHintProvider{
map[string][]TopologyHint{
"resource2": {
{
NUMANodeAffinity: NewTestBitMask(0, 3),
Preferred: false,
},
},
},
},
},
expected: TopologyHint{
NUMANodeAffinity: NewTestBitMask(0, 3),
Preferred: false,
},
},
{
name: "bestNonPreferredAffinityCount (3)",
hp: []HintProvider{
&mockHintProvider{
map[string][]TopologyHint{
"resource1": {
{
NUMANodeAffinity: NewTestBitMask(0, 1, 2, 3),
Preferred: false,
},
{
NUMANodeAffinity: NewTestBitMask(0, 1),
Preferred: false,
},
},
},
},
&mockHintProvider{
map[string][]TopologyHint{
"resource2": {
{
NUMANodeAffinity: NewTestBitMask(1, 2),
Preferred: false,
},
},
},
},
},
expected: TopologyHint{
NUMANodeAffinity: NewTestBitMask(1, 2),
Preferred: false,
},
},
{
name: "bestNonPreferredAffinityCount (4)",
hp: []HintProvider{
&mockHintProvider{
map[string][]TopologyHint{
"resource1": {
{
NUMANodeAffinity: NewTestBitMask(0, 1, 2, 3),
Preferred: false,
},
{
NUMANodeAffinity: NewTestBitMask(0, 1),
Preferred: false,
},
},
},
},
&mockHintProvider{
map[string][]TopologyHint{
"resource2": {
{
NUMANodeAffinity: NewTestBitMask(2, 3),
Preferred: false,
},
},
},
},
},
expected: TopologyHint{
NUMANodeAffinity: NewTestBitMask(2, 3),
Preferred: false,
},
},
{
name: "bestNonPreferredAffinityCount (5)",
hp: []HintProvider{
&mockHintProvider{
map[string][]TopologyHint{
"resource1": {
{
NUMANodeAffinity: NewTestBitMask(0, 1, 2, 3),
Preferred: false,
},
{
NUMANodeAffinity: NewTestBitMask(0, 1),
Preferred: false,
},
},
},
},
&mockHintProvider{
map[string][]TopologyHint{
"resource2": {
{
NUMANodeAffinity: NewTestBitMask(1, 2),
Preferred: false,
},
{
NUMANodeAffinity: NewTestBitMask(2, 3),
Preferred: false,
},
},
},
},
},
expected: TopologyHint{
NUMANodeAffinity: NewTestBitMask(1, 2),
Preferred: false,
},
},
{
name: "bestNonPreferredAffinityCount (6)",
hp: []HintProvider{
&mockHintProvider{
map[string][]TopologyHint{
"resource1": {
{
NUMANodeAffinity: NewTestBitMask(0, 1, 2, 3),
Preferred: false,
},
{
NUMANodeAffinity: NewTestBitMask(0, 1),
Preferred: false,
},
},
},
},
&mockHintProvider{
map[string][]TopologyHint{
"resource2": {
{
NUMANodeAffinity: NewTestBitMask(1, 2, 3),
Preferred: false,
},
{
NUMANodeAffinity: NewTestBitMask(1, 2),
Preferred: false,
},
{
NUMANodeAffinity: NewTestBitMask(1, 3),
Preferred: false,
},
{
NUMANodeAffinity: NewTestBitMask(2, 3),
Preferred: false,
},
},
},
},
},
expected: TopologyHint{
NUMANodeAffinity: NewTestBitMask(1, 2),
Preferred: false,
},
},
}
}