Search Mailing List Archives


Limit search to: Subject & Body Subject Author
Sort by: Reverse Sort
Limit to: All This Week Last Week This Month Last Month
Select Date Range     through    

[protege-discussion] Fwd: Protege 3.4.4 (579) 64-bit max heap size

Len Yabloko protege at ontospace.net
Sat Sep 4 08:18:11 PDT 2010


-----Original Message-----
From: Len Yabloko [mailto:protege at ontospace.net]
Sent: Saturday, September 4, 2010 11:17 AM
To: 'Timothy Redmond'
Subject: Re:  [protege-discussion] Protege 3.4.4 (579) 64-bit max heap     size

>
>On Sep 3, 2010, at 6:47 PM, Len Yabloko wrote:
>
>> Thank you for all resp onces. I spent much of the day trying to use  
>> all available memory for loading Protege frames. The best result I  
>> got was with
>> -XX:MaxPermSize=3000M, although I am not sure that was it. Here is  
>> the screenshot of Windows 7 task manager showing the big plato and  
>> then sudden jump at the end. Can someone explain it?
>
>I would guess  that such a MaxPermSize is not recommended - it is far  
>too large.  I am a bit uncertain as to what the issue is here.  Are  
>you getting OutOfMemory errors and if so what do they look like?  Do  
>they mention the PermGen space?  Simply not using the memory may not  
>be a problem.  Are you trying to reduce the amount of garbage  
>collection that is occurring?
>

I am not getting any errors. And yes- I am trying to maximize the throghput and avoid any CG overhead until the memory is completely exhausted. 

>Setting the initial heap size with -Xms will give java the hint that  
>you want to allocate more  space up front.  It is often recommended  
>that the Xmx and Xms get set at the same time for the reason that  
>Peris  gives below.  But this also makes java allocate space even if  
>it doesn't need it.

That would be fine in my case since no it is a client side. And I tried to set both to 6G, but jvm still allocates it as needed upto a cirtain limit under 4G. That alone I could understand as hard limit of 1/2 physical memory. But then at the end of loading data jvm memory jumps all the way as screeshot shows. Why not at the bgining?

>
>-Timothy
>
>
>
>
>>
>>
>>
>>>
>>> I would try setting the initial heap size, as well (-Xms6g).   
>>> Otherwise, the
>>> heap must be grown over the runtime in possibly many smaller  
>>> chunks, which
>>> can slow performance.
>>>
>>> -Peris
>>>
>>> On Fri, Sep 3, 2010 at 12:57 PM, Len Yabloko  
>>> <protege at ontospace.net> wrote:
>>>
>>>> Hello everyone,
>>>>
>>>> I am running latest Protege Frames version on Windows 7 64-bit and  
>>>> trying
>>>> to generate a very large frame ontology by loading data from MySQL  
>>>> database
>>>> using frame API. I have allocated 6G to jvm and checked  
>>>> Runtime.maxMemory()
>>>> to verify that is was correctly allocated. But for some reason  
>>>> when memory
>>>> reaches about 2.5G (according to Windows task manager) it stops  
>>>> allocating
>>>> and loading slows after about 15,000 frames loaded. I tried  
>>>> launching
>>>> Protege with -vmargs -Xmx6g using 64-bit jvm, but no luck so far.
>>>>
>>>> Can someone suggest what may be a cause of such behavior?
>>>>
>>>>
>>>> _______________________________________________
>>>> protege-discussion mailing list
>>>> protege-discussion at lists.stanford.edu
>>>> https://mailman.stanford.edu/mailman/listinfo/protege-discussion
>>>>
>>>> Instructions for unsubscribing:
>>>> http://protege.stanford.edu/doc/faq.html#01a.03
>>>>
>>>
>>>
>>>
>>> -- 
>>> “Ah, but a man's reach should exceed his grasp, or what's a heaven  
>>> for?”
>>>
>> <MaxPerm3000done.png>_______________________________________________
>> protege-discussion mailing list
>> protege-discussion at lists.stanford.edu
>> https://mailman.stanford.edu/mailman/listinfo/protege-discussion
>>
>> Instructions for unsubscribing: http://protege.stanford.edu/doc/faq.html#01a.03
>
>







More information about the protege-discussion mailing list