Chromium Code Reviews
|
|
Chromium Code Reviews
Issue
2018503002:
[system-health] Normalize the structure of memory metric values (Closed)
|
Created:
4 years, 7 months ago by petrcermak Modified:
4 years, 6 months ago CC:
catapult-reviews_chromium.org, tracing-review_chromium.org, eakuefner, benjhayden, perezju, sullivan Base URL:
git@github.com:catapult-project/catapult.git@master Target Ref:
refs/heads/master Project:
catapult Visibility:
Public. |
Description[system-health] Normalize the structure of memory metric values
As explained in the tracking bug
(https://github.com/catapult-project/catapult/issues/2363), this patch
normalizes the values reported by the memory metric as follows:
1. All values are renamed as follows:
Before: memory:<browser>:<process>:vmstats:
<component>:<size-property>
After: memory:<browser>:<process>:reported_by_os:system_memory:
<component>:<size-property>
Before: memory:<browser>:<process>:subsystem:gpu:
android_memtrack:<component>:<size-property>
After: memory:<browser>:<process>:reported_by_os:gpu_memory:
<component>:<size-property>
Before: memory:<browser>:<process>:subsystem:
<component>:<size-property>
(<component> != gpu:android_memtrack)
After: memory:<browser>:<process>:reported_by_chrome:
<component>:<size-property>
2. Top-level sums are added:
* memory:<browser>:<process>:reported_by_os:
proportional_resident_size
* memory:<browser>:<process>:reported_by_os:system_memory:
proportional_resident_size
(already reported as memory:<browser>:<process>:vmstats:
overall:pss)
* memory:<browser>:<process>:reported_by_os:gpu_memory:
proportional_resident_size
* memory:<browser>:<process>:reported_by_chrome:effective_size
4. 'pss' and 'private_dirty' are renamed to
'proportional_resident_size' and 'private_dirty_size'.
4. Process names are extended to 'browser_process',
'renderer_processes', 'gpu_process', 'ppapi_process' and
'all_processes'.
5. 'memtrack_pss' size property is renamed to
'proportional_resident_size' in the memory metric (but NOT in the
Trace Viewer UI).
6. The full cross product of {native_heap, java_heap, ashmem} and
{proportional_resident_size, private_dirty_size} is now reported:
* memory:<browser>:<process>:reported_by_os:system_memory:
native_heap:proportional_resident_size
* memory:<browser>:<process>:reported_by_os:system_memory:
java_heap:proportional_resident_size (NEW)
* memory:<browser>:<process>:reported_by_os:system_memory:
ashmem:proportional_resident_size
* memory:<browser>:<process>:reported_by_os:system_memory:
native_heap:private_dirty_size (NEW)
* memory:<browser>:<process>:reported_by_os:system_memory:
java_heap:private_dirty_size
* memory:<browser>:<process>:reported_by_os:system_memory:
ashmem:private_dirty_size (NEW)
This patch is an important step towards the metrics graphical model
(see https://goo.gl/uCI8Jh and https://goo.gl/gZp7SU).
BUG=chromium:617117
,catapult:#2363
Committed: https://chromium.googlesource.com/external/github.com/catapult-project/catapult/+/76449fc4e65a7e6b51d061d40539793a2e28371b
Patch Set 1 #
Total comments: 2
Patch Set 2 : Rebase #Patch Set 3 : Longer explicit names #Patch Set 4 : Exclude tracing from total reported_by_chrome values #
Messages
Total messages: 22 (12 generated)
Patchset #1 (id:1) has been deleted
Description was changed from ========== [system-health] Normalize the structure of memory metric values BUG=catapult:#2363 ========== to ========== [system-health] Normalize the structure of memory metric values As explained in the tracking bug (https://github.com/catapult-project/catapult/issues/2363), this patch normalizes the values reported by the memory metric as follows: 1. All values are renamed as follows: Before: memory:<browser>:<process>:vmstats: <component>:<size-property> After: memory:<browser>:<process>:reported_by_os:system: <component>:<size-property> Before: memory:<browser>:<process>:subsystem:gpu: android_memtrack:<component>:<size-property> After: memory:<browser>:<process>:reported_by_os:gpu: <component>:<size-property> Before: memory:<browser>:<process>:subsystem: <component>:<size-property> (<component> != gpu:android_memtrack) After: memory:<browser>:<process>:reported_by_chrome: <component>:<size-property> 2. Top-level sums are added: * memory:<browser>:<process>:reported_by_os:pss * memory:<browser>:<process>:reported_by_os:system:pss (already reported as memory:<browser>:<process>:vmstats: overall:pss) * memory:<browser>:<process>:reported_by_os:gpu:pss * memory:<browser>:<process>:reported_by_chrome:effective_size 3. 'memtrack_pss' size property is renamed to 'pss' in the memory metric (but NOT in the Trace Viewer UI). 4. The full cross product of {native_heap, java_heap, ashmem} and {pss, private_dirty} is now reported: * memory:<browser>:<process>:reported_by_os:native_heap:pss * memory:<browser>:<process>:reported_by_os:java_heap:pss (NEW) * memory:<browser>:<process>:reported_by_os:ashmem:pss * memory:<browser>:<process>:reported_by_os:native_heap: private_dirty (NEW) * memory:<browser>:<process>:reported_by_os:java_heap: private_dirty * memory:<browser>:<process>:reported_by_os:ashmem: private_dirty (NEW) This patch is an important step towards the metrics graphical model (see https://goo.gl/uCI8Jh and https://goo.gl/gZp7SU). BUG=catapult:#2363 ==========
petrcermak@chromium.org changed reviewers: + primiano@chromium.org
primiano: PTAL. I didn't do the four changes as separate patches because they require quite a bit of refactoring and go hand in hand. nednguyen, eakuefner, benjhayden, perezju: FYI. Thanks, Petr
nednguyen@google.com changed reviewers: + nednguyen@google.com
+Annie This will lead to a massive name update to the memory metrics on the dashboard. The consequence with this is you will have s.t like 460 old metrics + ~450 new metrics = 900 fields in the metrics drop down of your current tbm2 memory benchmark on the dashboard. Bisect will also messed up real bad. I think we should take this route instead: Create a new memoryMetricNew with all these name update. Clone existing memory benchmarks to use memoryMetricNew & let the old + new benchmarks running in parallel for a while Delete old memory benchmarks. Remove memoryMetric & rename memoryMetricNew to memoryMetric Remove
New names and code itself LGTM. Defer to Ned to the best approach to move forward. https://codereview.chromium.org/2018503002/diff/20001/tracing/tracing/metrics... File tracing/tracing/metrics/system_health/memory_metric.html (right): https://codereview.chromium.org/2018503002/diff/20001/tracing/tracing/metrics... tracing/tracing/metrics/system_health/memory_metric.html:252: * process_count value. 100 LOC to build a user friendly string. Might be user friendly, but not codereview friendly :P
Primiano: Thanks for the review. Ned: I was worried that changing the names could be a bit of an issue and also would like to get Annie's approval before going ahead. If we are to go the way you proposed (temporary separate memory metric), then I think it's very important that we do all value name changes in bulk (so that we're not in the same situation in a few weeks with another temporary metric). Probably the most important thing is to decide on https://bugs.chromium.org/p/chromium/issues/detail?id=615036#c1 (aggregation over browsers as well). One thing to mention is that there is a clear mapping from the old metrics to the new ones. So in theory, we could modify the past results. Also note that we won't be pushing 450 numbers anymore because we only report the _avg values. Furthermore, if we went with the proposal in https://bugs.chromium.org/p/chromium/issues/detail?id=610962#c18 (delete old non-avg memory values), we would be down to ~200 from both the old and the new value names (~100 if we renamed all past results). Thanks, Petr https://codereview.chromium.org/2018503002/diff/20001/tracing/tracing/metrics... File tracing/tracing/metrics/system_health/memory_metric.html (right): https://codereview.chromium.org/2018503002/diff/20001/tracing/tracing/metrics... tracing/tracing/metrics/system_health/memory_metric.html:252: * process_count value. On 2016/05/27 09:30:21, Primiano Tucci wrote: > 100 LOC to build a user friendly string. Might be user friendly, but not > codereview friendly :P True. I'm super grateful that the descriptions are in English though. It would be much more difficult if we did this in a language that has full declension (https://en.wikipedia.org/wiki/Czech_declension).
On 2016/05/27 09:47:52, petrcermak wrote: > One thing to mention is that there is a clear mapping from the old metrics to > the new ones. So in theory, we could modify the past results. I'd keep it simple and just drop all the old data, IIRC we don't have anything particularly critical yet in top_10_mobile_tbmv2. What other changes we plan in future? > https://codereview.chromium.org/2018503002/diff/20001/tracing/tracing/metrics... > tracing/tracing/metrics/system_health/memory_metric.html:252: * process_count > value. > On 2016/05/27 09:30:21, Primiano Tucci wrote: > > 100 LOC to build a user friendly string. Might be user friendly, but not > > codereview friendly :P > > True. I'm super grateful that the descriptions are in English though. It would > be much more difficult if we did this in a language that has full declension > (https://en.wikipedia.org/wiki/Czech_declension). Suddenly I feel all those hours studying Latin in high school are worth something :D
On 2016/05/27 09:53:42, Primiano Tucci wrote: > On 2016/05/27 09:47:52, petrcermak wrote: > > One thing to mention is that there is a clear mapping from the old metrics to > > the new ones. So in theory, we could modify the past results. > > I'd keep it simple and just drop all the old data, IIRC we don't have anything > particularly critical yet in top_10_mobile_tbmv2. > What other changes we plan in future? > > > > https://codereview.chromium.org/2018503002/diff/20001/tracing/tracing/metrics... > > tracing/tracing/metrics/system_health/memory_metric.html:252: * process_count > > value. > > On 2016/05/27 09:30:21, Primiano Tucci wrote: > > > 100 LOC to build a user friendly string. Might be user friendly, but not > > > codereview friendly :P > > > > True. I'm super grateful that the descriptions are in English though. It would > > be much more difficult if we did this in a language that has full declension > > (https://en.wikipedia.org/wiki/Czech_declension). > Suddenly I feel all those hours studying Latin in high school are worth > something :D I am fine with doing this mass name migration if you are ok with ignoring the regression in one benchmark run. :-) Annie probably can help with removing the old names.
lgtm
Description was changed from ========== [system-health] Normalize the structure of memory metric values As explained in the tracking bug (https://github.com/catapult-project/catapult/issues/2363), this patch normalizes the values reported by the memory metric as follows: 1. All values are renamed as follows: Before: memory:<browser>:<process>:vmstats: <component>:<size-property> After: memory:<browser>:<process>:reported_by_os:system: <component>:<size-property> Before: memory:<browser>:<process>:subsystem:gpu: android_memtrack:<component>:<size-property> After: memory:<browser>:<process>:reported_by_os:gpu: <component>:<size-property> Before: memory:<browser>:<process>:subsystem: <component>:<size-property> (<component> != gpu:android_memtrack) After: memory:<browser>:<process>:reported_by_chrome: <component>:<size-property> 2. Top-level sums are added: * memory:<browser>:<process>:reported_by_os:pss * memory:<browser>:<process>:reported_by_os:system:pss (already reported as memory:<browser>:<process>:vmstats: overall:pss) * memory:<browser>:<process>:reported_by_os:gpu:pss * memory:<browser>:<process>:reported_by_chrome:effective_size 3. 'memtrack_pss' size property is renamed to 'pss' in the memory metric (but NOT in the Trace Viewer UI). 4. The full cross product of {native_heap, java_heap, ashmem} and {pss, private_dirty} is now reported: * memory:<browser>:<process>:reported_by_os:native_heap:pss * memory:<browser>:<process>:reported_by_os:java_heap:pss (NEW) * memory:<browser>:<process>:reported_by_os:ashmem:pss * memory:<browser>:<process>:reported_by_os:native_heap: private_dirty (NEW) * memory:<browser>:<process>:reported_by_os:java_heap: private_dirty * memory:<browser>:<process>:reported_by_os:ashmem: private_dirty (NEW) This patch is an important step towards the metrics graphical model (see https://goo.gl/uCI8Jh and https://goo.gl/gZp7SU). BUG=catapult:#2363 ========== to ========== [system-health] Normalize the structure of memory metric values As explained in the tracking bug (https://github.com/catapult-project/catapult/issues/2363), this patch normalizes the values reported by the memory metric as follows: 1. All values are renamed as follows: Before: memory:<browser>:<process>:vmstats: <component>:<size-property> After: memory:<browser>:<process>:reported_by_os:system_memory: <component>:<size-property> Before: memory:<browser>:<process>:subsystem:gpu: android_memtrack:<component>:<size-property> After: memory:<browser>:<process>:reported_by_os:gpu_memory: <component>:<size-property> Before: memory:<browser>:<process>:subsystem: <component>:<size-property> (<component> != gpu:android_memtrack) After: memory:<browser>:<process>:reported_by_chrome: <component>:<size-property> 2. Top-level sums are added: * memory:<browser>:<process>:reported_by_os: proportional_resident_size * memory:<browser>:<process>:reported_by_os:system_memory: proportional_resident_size (already reported as memory:<browser>:<process>:vmstats: overall:pss) * memory:<browser>:<process>:reported_by_os:gpu_memory: proportional_resident_size * memory:<browser>:<process>:reported_by_chrome:effective_size 3. 'memtrack_pss' size property is renamed to 'proportional_resident_size' in the memory metric (but NOT in the Trace Viewer UI). 4. The full cross product of {native_heap, java_heap, ashmem} and {proportional_resident_size, private_dirty_size} is now reported: * memory:<browser>:<process>:reported_by_os:system_memory: native_heap:proportional_resident_size * memory:<browser>:<process>:reported_by_os:system_memory: java_heap:proportional_resident_size (NEW) * memory:<browser>:<process>:reported_by_os:system_memory: ashmem:proportional_resident_size * memory:<browser>:<process>:reported_by_os:system_memory: native_heap:private_dirty_size (NEW) * memory:<browser>:<process>:reported_by_os:system_memory: java_heap:private_dirty_size * memory:<browser>:<process>:reported_by_os:system_memory: ashmem:private_dirty_size (NEW) This patch is an important step towards the metrics graphical model (see https://goo.gl/uCI8Jh and https://goo.gl/gZp7SU). BUG=catapult:#2363,chromium:617117 ==========
Description was changed from ========== [system-health] Normalize the structure of memory metric values As explained in the tracking bug (https://github.com/catapult-project/catapult/issues/2363), this patch normalizes the values reported by the memory metric as follows: 1. All values are renamed as follows: Before: memory:<browser>:<process>:vmstats: <component>:<size-property> After: memory:<browser>:<process>:reported_by_os:system_memory: <component>:<size-property> Before: memory:<browser>:<process>:subsystem:gpu: android_memtrack:<component>:<size-property> After: memory:<browser>:<process>:reported_by_os:gpu_memory: <component>:<size-property> Before: memory:<browser>:<process>:subsystem: <component>:<size-property> (<component> != gpu:android_memtrack) After: memory:<browser>:<process>:reported_by_chrome: <component>:<size-property> 2. Top-level sums are added: * memory:<browser>:<process>:reported_by_os: proportional_resident_size * memory:<browser>:<process>:reported_by_os:system_memory: proportional_resident_size (already reported as memory:<browser>:<process>:vmstats: overall:pss) * memory:<browser>:<process>:reported_by_os:gpu_memory: proportional_resident_size * memory:<browser>:<process>:reported_by_chrome:effective_size 3. 'memtrack_pss' size property is renamed to 'proportional_resident_size' in the memory metric (but NOT in the Trace Viewer UI). 4. The full cross product of {native_heap, java_heap, ashmem} and {proportional_resident_size, private_dirty_size} is now reported: * memory:<browser>:<process>:reported_by_os:system_memory: native_heap:proportional_resident_size * memory:<browser>:<process>:reported_by_os:system_memory: java_heap:proportional_resident_size (NEW) * memory:<browser>:<process>:reported_by_os:system_memory: ashmem:proportional_resident_size * memory:<browser>:<process>:reported_by_os:system_memory: native_heap:private_dirty_size (NEW) * memory:<browser>:<process>:reported_by_os:system_memory: java_heap:private_dirty_size * memory:<browser>:<process>:reported_by_os:system_memory: ashmem:private_dirty_size (NEW) This patch is an important step towards the metrics graphical model (see https://goo.gl/uCI8Jh and https://goo.gl/gZp7SU). BUG=catapult:#2363,chromium:617117 ========== to ========== [system-health] Normalize the structure of memory metric values As explained in the tracking bug (https://github.com/catapult-project/catapult/issues/2363), this patch normalizes the values reported by the memory metric as follows: 1. All values are renamed as follows: Before: memory:<browser>:<process>:vmstats: <component>:<size-property> After: memory:<browser>:<process>:reported_by_os:system_memory: <component>:<size-property> Before: memory:<browser>:<process>:subsystem:gpu: android_memtrack:<component>:<size-property> After: memory:<browser>:<process>:reported_by_os:gpu_memory: <component>:<size-property> Before: memory:<browser>:<process>:subsystem: <component>:<size-property> (<component> != gpu:android_memtrack) After: memory:<browser>:<process>:reported_by_chrome: <component>:<size-property> 2. Top-level sums are added: * memory:<browser>:<process>:reported_by_os: proportional_resident_size * memory:<browser>:<process>:reported_by_os:system_memory: proportional_resident_size (already reported as memory:<browser>:<process>:vmstats: overall:pss) * memory:<browser>:<process>:reported_by_os:gpu_memory: proportional_resident_size * memory:<browser>:<process>:reported_by_chrome:effective_size 3. 'memtrack_pss' size property is renamed to 'proportional_resident_size' in the memory metric (but NOT in the Trace Viewer UI). 4. The full cross product of {native_heap, java_heap, ashmem} and {proportional_resident_size, private_dirty_size} is now reported: * memory:<browser>:<process>:reported_by_os:system_memory: native_heap:proportional_resident_size * memory:<browser>:<process>:reported_by_os:system_memory: java_heap:proportional_resident_size (NEW) * memory:<browser>:<process>:reported_by_os:system_memory: ashmem:proportional_resident_size * memory:<browser>:<process>:reported_by_os:system_memory: native_heap:private_dirty_size (NEW) * memory:<browser>:<process>:reported_by_os:system_memory: java_heap:private_dirty_size * memory:<browser>:<process>:reported_by_os:system_memory: ashmem:private_dirty_size (NEW) This patch is an important step towards the metrics graphical model (see https://goo.gl/uCI8Jh and https://goo.gl/gZp7SU). BUG=chromium:617117,catapult:#2363 ==========
Patchset #3 (id:60001) has been deleted
Description was changed from ========== [system-health] Normalize the structure of memory metric values As explained in the tracking bug (https://github.com/catapult-project/catapult/issues/2363), this patch normalizes the values reported by the memory metric as follows: 1. All values are renamed as follows: Before: memory:<browser>:<process>:vmstats: <component>:<size-property> After: memory:<browser>:<process>:reported_by_os:system_memory: <component>:<size-property> Before: memory:<browser>:<process>:subsystem:gpu: android_memtrack:<component>:<size-property> After: memory:<browser>:<process>:reported_by_os:gpu_memory: <component>:<size-property> Before: memory:<browser>:<process>:subsystem: <component>:<size-property> (<component> != gpu:android_memtrack) After: memory:<browser>:<process>:reported_by_chrome: <component>:<size-property> 2. Top-level sums are added: * memory:<browser>:<process>:reported_by_os: proportional_resident_size * memory:<browser>:<process>:reported_by_os:system_memory: proportional_resident_size (already reported as memory:<browser>:<process>:vmstats: overall:pss) * memory:<browser>:<process>:reported_by_os:gpu_memory: proportional_resident_size * memory:<browser>:<process>:reported_by_chrome:effective_size 3. 'memtrack_pss' size property is renamed to 'proportional_resident_size' in the memory metric (but NOT in the Trace Viewer UI). 4. The full cross product of {native_heap, java_heap, ashmem} and {proportional_resident_size, private_dirty_size} is now reported: * memory:<browser>:<process>:reported_by_os:system_memory: native_heap:proportional_resident_size * memory:<browser>:<process>:reported_by_os:system_memory: java_heap:proportional_resident_size (NEW) * memory:<browser>:<process>:reported_by_os:system_memory: ashmem:proportional_resident_size * memory:<browser>:<process>:reported_by_os:system_memory: native_heap:private_dirty_size (NEW) * memory:<browser>:<process>:reported_by_os:system_memory: java_heap:private_dirty_size * memory:<browser>:<process>:reported_by_os:system_memory: ashmem:private_dirty_size (NEW) This patch is an important step towards the metrics graphical model (see https://goo.gl/uCI8Jh and https://goo.gl/gZp7SU). BUG=chromium:617117,catapult:#2363 ========== to ========== [system-health] Normalize the structure of memory metric values As explained in the tracking bug (https://github.com/catapult-project/catapult/issues/2363), this patch normalizes the values reported by the memory metric as follows: 1. All values are renamed as follows: Before: memory:<browser>:<process>:vmstats: <component>:<size-property> After: memory:<browser>:<process>:reported_by_os:system_memory: <component>:<size-property> Before: memory:<browser>:<process>:subsystem:gpu: android_memtrack:<component>:<size-property> After: memory:<browser>:<process>:reported_by_os:gpu_memory: <component>:<size-property> Before: memory:<browser>:<process>:subsystem: <component>:<size-property> (<component> != gpu:android_memtrack) After: memory:<browser>:<process>:reported_by_chrome: <component>:<size-property> 2. Top-level sums are added: * memory:<browser>:<process>:reported_by_os: proportional_resident_size * memory:<browser>:<process>:reported_by_os:system_memory: proportional_resident_size (already reported as memory:<browser>:<process>:vmstats: overall:pss) * memory:<browser>:<process>:reported_by_os:gpu_memory: proportional_resident_size * memory:<browser>:<process>:reported_by_chrome:effective_size 4. 'pss' and 'private_dirty' are renamed to 'proportional_resident_size' and 'private_dirty_size'. 4. Process names are extended to 'browser_process', 'renderer_processes', 'gpu_process', 'ppapi_process' and 'all_processes'. 5. 'memtrack_pss' size property is renamed to 'proportional_resident_size' in the memory metric (but NOT in the Trace Viewer UI). 6. The full cross product of {native_heap, java_heap, ashmem} and {proportional_resident_size, private_dirty_size} is now reported: * memory:<browser>:<process>:reported_by_os:system_memory: native_heap:proportional_resident_size * memory:<browser>:<process>:reported_by_os:system_memory: java_heap:proportional_resident_size (NEW) * memory:<browser>:<process>:reported_by_os:system_memory: ashmem:proportional_resident_size * memory:<browser>:<process>:reported_by_os:system_memory: native_heap:private_dirty_size (NEW) * memory:<browser>:<process>:reported_by_os:system_memory: java_heap:private_dirty_size * memory:<browser>:<process>:reported_by_os:system_memory: ashmem:private_dirty_size (NEW) This patch is an important step towards the metrics graphical model (see https://goo.gl/uCI8Jh and https://goo.gl/gZp7SU). BUG=chromium:617117,catapult:#2363 ==========
Patchset #4 (id:100001) has been deleted
FYI this patch has just been updated to its final version and will land tomorrow.
The CQ bit was checked by petrcermak@chromium.org
The patchset sent to the CQ was uploaded after l-g-t-m from primiano@chromium.org, nednguyen@google.com Link to the patchset: https://codereview.chromium.org/2018503002/#ps120001 (title: "Exclude tracing from total reported_by_chrome values")
CQ is trying da patch. Follow status at https://chromium-cq-status.appspot.com/patch-status/2018503002/120001
Message was sent while issue was closed.
Description was changed from ========== [system-health] Normalize the structure of memory metric values As explained in the tracking bug (https://github.com/catapult-project/catapult/issues/2363), this patch normalizes the values reported by the memory metric as follows: 1. All values are renamed as follows: Before: memory:<browser>:<process>:vmstats: <component>:<size-property> After: memory:<browser>:<process>:reported_by_os:system_memory: <component>:<size-property> Before: memory:<browser>:<process>:subsystem:gpu: android_memtrack:<component>:<size-property> After: memory:<browser>:<process>:reported_by_os:gpu_memory: <component>:<size-property> Before: memory:<browser>:<process>:subsystem: <component>:<size-property> (<component> != gpu:android_memtrack) After: memory:<browser>:<process>:reported_by_chrome: <component>:<size-property> 2. Top-level sums are added: * memory:<browser>:<process>:reported_by_os: proportional_resident_size * memory:<browser>:<process>:reported_by_os:system_memory: proportional_resident_size (already reported as memory:<browser>:<process>:vmstats: overall:pss) * memory:<browser>:<process>:reported_by_os:gpu_memory: proportional_resident_size * memory:<browser>:<process>:reported_by_chrome:effective_size 4. 'pss' and 'private_dirty' are renamed to 'proportional_resident_size' and 'private_dirty_size'. 4. Process names are extended to 'browser_process', 'renderer_processes', 'gpu_process', 'ppapi_process' and 'all_processes'. 5. 'memtrack_pss' size property is renamed to 'proportional_resident_size' in the memory metric (but NOT in the Trace Viewer UI). 6. The full cross product of {native_heap, java_heap, ashmem} and {proportional_resident_size, private_dirty_size} is now reported: * memory:<browser>:<process>:reported_by_os:system_memory: native_heap:proportional_resident_size * memory:<browser>:<process>:reported_by_os:system_memory: java_heap:proportional_resident_size (NEW) * memory:<browser>:<process>:reported_by_os:system_memory: ashmem:proportional_resident_size * memory:<browser>:<process>:reported_by_os:system_memory: native_heap:private_dirty_size (NEW) * memory:<browser>:<process>:reported_by_os:system_memory: java_heap:private_dirty_size * memory:<browser>:<process>:reported_by_os:system_memory: ashmem:private_dirty_size (NEW) This patch is an important step towards the metrics graphical model (see https://goo.gl/uCI8Jh and https://goo.gl/gZp7SU). BUG=chromium:617117,catapult:#2363 ========== to ========== [system-health] Normalize the structure of memory metric values As explained in the tracking bug (https://github.com/catapult-project/catapult/issues/2363), this patch normalizes the values reported by the memory metric as follows: 1. All values are renamed as follows: Before: memory:<browser>:<process>:vmstats: <component>:<size-property> After: memory:<browser>:<process>:reported_by_os:system_memory: <component>:<size-property> Before: memory:<browser>:<process>:subsystem:gpu: android_memtrack:<component>:<size-property> After: memory:<browser>:<process>:reported_by_os:gpu_memory: <component>:<size-property> Before: memory:<browser>:<process>:subsystem: <component>:<size-property> (<component> != gpu:android_memtrack) After: memory:<browser>:<process>:reported_by_chrome: <component>:<size-property> 2. Top-level sums are added: * memory:<browser>:<process>:reported_by_os: proportional_resident_size * memory:<browser>:<process>:reported_by_os:system_memory: proportional_resident_size (already reported as memory:<browser>:<process>:vmstats: overall:pss) * memory:<browser>:<process>:reported_by_os:gpu_memory: proportional_resident_size * memory:<browser>:<process>:reported_by_chrome:effective_size 4. 'pss' and 'private_dirty' are renamed to 'proportional_resident_size' and 'private_dirty_size'. 4. Process names are extended to 'browser_process', 'renderer_processes', 'gpu_process', 'ppapi_process' and 'all_processes'. 5. 'memtrack_pss' size property is renamed to 'proportional_resident_size' in the memory metric (but NOT in the Trace Viewer UI). 6. The full cross product of {native_heap, java_heap, ashmem} and {proportional_resident_size, private_dirty_size} is now reported: * memory:<browser>:<process>:reported_by_os:system_memory: native_heap:proportional_resident_size * memory:<browser>:<process>:reported_by_os:system_memory: java_heap:proportional_resident_size (NEW) * memory:<browser>:<process>:reported_by_os:system_memory: ashmem:proportional_resident_size * memory:<browser>:<process>:reported_by_os:system_memory: native_heap:private_dirty_size (NEW) * memory:<browser>:<process>:reported_by_os:system_memory: java_heap:private_dirty_size * memory:<browser>:<process>:reported_by_os:system_memory: ashmem:private_dirty_size (NEW) This patch is an important step towards the metrics graphical model (see https://goo.gl/uCI8Jh and https://goo.gl/gZp7SU). BUG=chromium:617117,catapult:#2363 Committed: https://chromium.googlesource.com/external/github.com/catapult-project/catapu... ==========
Message was sent while issue was closed.
Committed patchset #4 (id:120001) as https://chromium.googlesource.com/external/github.com/catapult-project/catapu... |
|||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
