Re: Review Request 43686: Added allocator benchmark using labeled resources.

[Neil Conway]

---
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/43686/
---

(Updated March 2, 2016, 6:20 a.m.)


Review request for mesos, Joris Van Remoortere and Michael Park.


Changes
---

Address code review comments.


Bugs: MESOS-4691
https://issues.apache.org/jira/browse/MESOS-4691


Repository: mesos


Description
---

This reveals that when the cluster contains many reservations with
distinct labels, allocator performance slows down dramatically. A
short-term fix for this problem will be introduced shortly.


Diffs (updated)
-

  src/tests/hierarchical_allocator_tests.cpp 
5f771f02db9bd098f3cd36730cd84bf2f5e87a33 

Diff: https://reviews.apache.org/r/43686/diff/


Testing
---

make check

FYI, results on my laptop:

_Original benchmark (unlabeled resources)_
[ RUN  ] HierarchicalAllocator_BENCHMARK_Test.DeclineOffers
Using 2000 slaves and 200 frameworks
round 0 allocate took 2.028175secs to make 200 offers
round 1 allocate took 2.006791secs to make 200 offers
round 2 allocate took 2.033723secs to make 200 offers
round 3 allocate took 2.017508secs to make 200 offers
round 4 allocate took 2.037235secs to make 200 offers
round 5 allocate took 2.054095secs to make 200 offers
round 6 allocate took 2.048884secs to make 200 offers
round 7 allocate took 2.044252secs to make 200 offers
round 8 allocate took 2.060256secs to make 200 offers
round 9 allocate took 2.07121secs to make 200 offers
round 10 allocate took 2.066261secs to make 200 offers
round 11 allocate took 2.034805secs to make 200 offers
round 12 allocate took 2.053705secs to make 200 offers
round 13 allocate took 2.042106secs to make 200 offers
round 14 allocate took 2.082704secs to make 200 offers


Thanks,

Neil Conway

Re: Review Request 43686: Added allocator benchmark using labeled resources.

[Joris Van Remoortere]

---
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/43686/#review121581
---


Fix it, then Ship it!





src/tests/hierarchical_allocator_tests.cpp (lines 2610 - 2620)


let's use a loop, and add a comment explaining that we're trying to force 
the N^2 labels comparison algorithm by having so many labels.



src/tests/hierarchical_allocator_tests.cpp (lines 2633 - 2634)


Can we add a TODO to factor this test out by parameterizing the 
DeclineFilter test of which it is a copy?


- Joris Van Remoortere


On March 2, 2016, 2:21 a.m., Neil Conway wrote:
> 
> ---
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/43686/
> ---
> 
> (Updated March 2, 2016, 2:21 a.m.)
> 
> 
> Review request for mesos, Joris Van Remoortere and Michael Park.
> 
> 
> Bugs: MESOS-4691
> https://issues.apache.org/jira/browse/MESOS-4691
> 
> 
> Repository: mesos
> 
> 
> Description
> ---
> 
> This reveals that when the cluster contains many reservations with
> distinct labels, allocator performance slows down dramatically. A
> short-term fix for this problem will be introduced shortly.
> 
> 
> Diffs
> -
> 
>   src/tests/hierarchical_allocator_tests.cpp 
> 5f771f02db9bd098f3cd36730cd84bf2f5e87a33 
> 
> Diff: https://reviews.apache.org/r/43686/diff/
> 
> 
> Testing
> ---
> 
> make check
> 
> FYI, results on my laptop:
> 
> _Original benchmark (unlabeled resources)_
> [ RUN  ] HierarchicalAllocator_BENCHMARK_Test.DeclineOffers
> Using 2000 slaves and 200 frameworks
> round 0 allocate took 2.028175secs to make 200 offers
> round 1 allocate took 2.006791secs to make 200 offers
> round 2 allocate took 2.033723secs to make 200 offers
> round 3 allocate took 2.017508secs to make 200 offers
> round 4 allocate took 2.037235secs to make 200 offers
> round 5 allocate took 2.054095secs to make 200 offers
> round 6 allocate took 2.048884secs to make 200 offers
> round 7 allocate took 2.044252secs to make 200 offers
> round 8 allocate took 2.060256secs to make 200 offers
> round 9 allocate took 2.07121secs to make 200 offers
> round 10 allocate took 2.066261secs to make 200 offers
> round 11 allocate took 2.034805secs to make 200 offers
> round 12 allocate took 2.053705secs to make 200 offers
> round 13 allocate took 2.042106secs to make 200 offers
> round 14 allocate took 2.082704secs to make 200 offers
> 
> 
> Thanks,
> 
> Neil Conway
> 
>

Re: Review Request 43686: Added allocator benchmark using labeled resources.

[Neil Conway]

---
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/43686/
---

(Updated March 2, 2016, 2:21 a.m.)


Review request for mesos, Joris Van Remoortere and Michael Park.


Changes
---

Changed benchmark to isolate allocator bottleneck


Summary (updated)
-

Added allocator benchmark using labeled resources.


Bugs: MESOS-4691
https://issues.apache.org/jira/browse/MESOS-4691


Repository: mesos


Description (updated)
---

This reveals that when the cluster contains many reservations with
distinct labels, allocator performance slows down dramatically. A
short-term fix for this problem will be introduced shortly.


Diffs (updated)
-

  src/tests/hierarchical_allocator_tests.cpp 
5f771f02db9bd098f3cd36730cd84bf2f5e87a33 

Diff: https://reviews.apache.org/r/43686/diff/


Testing (updated)
---

make check

FYI, results on my laptop:

_Original benchmark (unlabeled resources)_
[ RUN  ] HierarchicalAllocator_BENCHMARK_Test.DeclineOffers
Using 2000 slaves and 200 frameworks
round 0 allocate took 2.028175secs to make 200 offers
round 1 allocate took 2.006791secs to make 200 offers
round 2 allocate took 2.033723secs to make 200 offers
round 3 allocate took 2.017508secs to make 200 offers
round 4 allocate took 2.037235secs to make 200 offers
round 5 allocate took 2.054095secs to make 200 offers
round 6 allocate took 2.048884secs to make 200 offers
round 7 allocate took 2.044252secs to make 200 offers
round 8 allocate took 2.060256secs to make 200 offers
round 9 allocate took 2.07121secs to make 200 offers
round 10 allocate took 2.066261secs to make 200 offers
round 11 allocate took 2.034805secs to make 200 offers
round 12 allocate took 2.053705secs to make 200 offers
round 13 allocate took 2.042106secs to make 200 offers
round 14 allocate took 2.082704secs to make 200 offers


Thanks,

Neil Conway