aurora-reviews mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Maxim Khutornenko" <ma...@apache.org>
Subject Re: Review Request 18526: Add support for slaveHosts set in TaskQuery.
Date Fri, 28 Feb 2014 22:53:01 GMT

-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/18526/#review35861
-----------------------------------------------------------



src/main/java/org/apache/aurora/scheduler/storage/mem/MemTaskStore.java
<https://reviews.apache.org/r/18526/#comment66647>

    That would create an empty Set<IJobKey> in case jobkey is missing whereas I want
Optional<Set<IJobKey>>. 
    
    This code will be refactored shortly when I get to AURORA-235.



src/main/java/org/apache/aurora/scheduler/storage/mem/MemTaskStore.java
<https://reviews.apache.org/r/18526/#comment66654>

    Unfortunately, I could not find anything that could be converted into getAll() on a multimap.
The filter was the best I could find. Yes, it does iterate over keys to apply filtering predicate
as evident from the sources.
    
    I put .filter against a looped get() for a multimap size of ~500k (10k keys, up to 100
values per key) and, as expected, the perf of a looped get() degraded quickly with the increasing
query key-set size:
    
    Query set size   Perf
    1-10             .filter 50% slower
    100              .filter 20% slower
    1000             .filter 250% faster
    
    However, a looped get() outperformed .filter for anything below 100 keys. Since the majority
of our queries are well under 10 keys using a looped get() seems to be a better alternative.
Converting to it now.


- Maxim Khutornenko


On Feb. 28, 2014, 1:18 a.m., Maxim Khutornenko wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/18526/
> -----------------------------------------------------------
> 
> (Updated Feb. 28, 2014, 1:18 a.m.)
> 
> 
> Review request for Aurora, Kevin Sweeney and Bill Farner.
> 
> 
> Bugs: AURORA-232
>     https://issues.apache.org/jira/browse/AURORA-232
> 
> 
> Repository: aurora
> 
> 
> Description
> -------
> 
> The slaveHosts set-based field will eventually replace a string field. 
> All internal queries are converted to use the new field. The old one is still supported
for queries coming from the web (until client side is refactored).
> 
> 
> Diffs
> -----
> 
>   src/main/java/org/apache/aurora/scheduler/base/Query.java b9f207c740362fafff0257988f52b6179025d58d

>   src/main/java/org/apache/aurora/scheduler/http/Maintenance.java 7337044bacd052c516e4d78a2993946d472ef91c

>   src/main/java/org/apache/aurora/scheduler/storage/mem/MemTaskStore.java d1ab503e4edac86afcb8884a074a87b7536de3f7

>   src/main/python/apache/aurora/client/api/sla.py 131c357d60fd00740b51055f555d56c599124d15

>   src/main/thrift/org/apache/aurora/gen/api.thrift cd60f47bf34b4a634004e2ad9eadad37aa1556bb

>   src/test/java/org/apache/aurora/scheduler/storage/mem/MemTaskStoreTest.java 884f589a2cbea918ecbfedf457f42d7cc9254c95

>   src/test/resources/org/apache/aurora/gen/api.thrift.md5 fafb5100443482e662db453429c5259f2ab80ae5

> 
> Diff: https://reviews.apache.org/r/18526/diff/
> 
> 
> Testing
> -------
> 
> ./build-support/jenkins/build.sh 
> 
> 
> Thanks,
> 
> Maxim Khutornenko
> 
>


Mime
  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message