Home | About | Sematext search-lucene.com search-hadoop.com
 Search Hadoop and all its subprojects:

Switch to Threaded View
HBase, mail # dev - HFilePerformanceEvaluation: Nothing follows logic


Copy link to this message
-
Re: HFilePerformanceEvaluation: Nothing follows logic
Jean-Marc Spaggiari 2013-02-27, 14:39
Ok. I have done some tests.

And never came above a 0.0001% difference between the number of calls
expected to be done vs the number of calls really done.

So I don't think we need to worry about that and simply return when
scanner.next returns false.

JM

2013/2/27 Jean-Marc Spaggiari <[EMAIL PROTECTED]>:
> Thanks for your feedback.
>
> I will modify the patch to go with the 3rd option instead of the 2nd one.
>
> JM
>
> 2013/2/27 Ted <[EMAIL PROTECTED]>:
>> Thanks for your quick action.
>>
>> I think the third option is better - we get performance comparison without repeating the test.
>>
>> Cheers
>>
>> On Feb 27, 2013, at 5:18 AM, Jean-Marc Spaggiari <[EMAIL PROTECTED]> wrote:
>>
>>> Just opened HBASE-7953. Can you please take a look at it?
>>>
>>> There was 3 ways to fix that.
>>>
>>> 1) Simply exit the loop when scanner.next() return false
>>> 2) Exit the loop and restart the test
>>> 3) Count the number of scanner.next() effectively done and measure
>>> performances based on that and not on the ROW_COUNT value.
>>>
>>> I chose the 2nd one, but maybe there is some preferences for the 3rd one?
>>>
>>> JM
>>>
>>> 2013/2/27 Jean-Marc Spaggiari <[EMAIL PROTECTED]>:
>>>> Yep, I will open one and submit (and not commit ;) ) a patch.
>>>>
>>>> Thanks.
>>>>
>>>> JM
>>>>
>>>> 2013/2/27 Ted <[EMAIL PROTECTED]>:
>>>>> Looks like a bug.
>>>>>
>>>>> Mind opening Jira ?
>>>>>
>>>>> Thanks
>>>>>
>>>>> On Feb 27, 2013, at 4:32 AM, Jean-Marc Spaggiari <[EMAIL PROTECTED]> wrote:
>>>>>
>>>>>> Hi,
>>>>>>
>>>>>> In HFilePerformanceEvaluation, there is 2 places where we are doing
>>>>>> scanner.next() but still trying to read the key and/or the value. This
>>>>>> is failing. I'm wondering if there is any logic behind that or if I
>>>>>> can fix it.
>>>>>>
>>>>>> JM
>>>>>>
>>>>>>
>>>>>>     for (int ii = 0; ii < 30; ii++) {
>>>>>>       if (!scanner.next()) {
>>>>>>         System.out.println("NOTHING FOLLOWS");
>>>>>>       }
>>>>>>       scanner.getKey();
>>>>>>       scanner.getValue();
>>>>>>     }