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

Switch to Threaded View
HDFS >> mail # user >> Need help regarding HDFS-RAID


Copy link to this message
-
Re: Need help regarding HDFS-RAID
Hi Dhruba,

Thanks for the pointer. I'm going to try and pull this code into our internal 20-ish distro. Would you object if I make a contribution of that result if it is successful?
Best regards,
    - Andy

Problems worthy of attack prove their worth by hitting back. - Piet Hein (via Tom White)

>________________________________
>From: Dhruba Borthakur <[EMAIL PROTECTED]>
>To: Andrew Purtell <[EMAIL PROTECTED]>
>Cc: "[EMAIL PROTECTED]" <[EMAIL PROTECTED]>
>Sent: Tuesday, September 20, 2011 2:18 AM
>Subject: Re: Need help regarding HDFS-RAID
>
>
>Hi andy,
>
>
>we do run a version of HDFS RAID that is backported from Apache trunk to a 0.20 based release. Our code is in https://github.com/facebook/hadoop-20-warehouse/tree/master/src/contrib/raid
>But I do not have an elegant way to contribute this code to Apache 0.20.2xx.x. 
>
>
>thanks,
>dhruba
>
>
>On Sat, Sep 17, 2011 at 9:16 AM, Andrew Purtell <[EMAIL PROTECTED]> wrote:
>
>Hi Dhruba,
>>
>>
>>Would you consider a contribution of this to branch-0.20-security aka 0.20.2xx.x?
>>
>>
>>If I am mistaken and you do not have a 0.22-ish HDFS RAID backported to an 0.20-ish platform, please disregard.
>>
>>
>>Best regards,
>>
>>
>>    - Andy
>>
>>Problems worthy of attack prove their worth by hitting back. - Piet Hein (via Tom White)
>>
>>
>>>________________________________
>>>From: Dhruba Borthakur <[EMAIL PROTECTED]>
>>>To: [EMAIL PROTECTED]; Andrew Purtell <[EMAIL PROTECTED]>
>>>Sent: Thursday, September 15, 2011 10:14 AM
>>>
>>>Subject: Re: Need help regarding HDFS-RAID
>>>
>>>
>>>
>>>That's right Andy. 0.22+. We are running a HDFS-RAID code base that is pretty close to what is available in Apache hdfs trunk.
>>>
>>>
>>>-dhruba
>>>
>>>
>>>On Thu, Sep 15, 2011 at 10:08 AM, Andrew Purtell <[EMAIL PROTECTED]> wrote:
>>>
>>>But that is the HDFS RAID effectively in 0.22+, not 0.21, right Dhruba?
>>>>
>>>> 
>>>>Best regards,
>>>>
>>>>
>>>>       - Andy
>>>>
>>>>Problems worthy of attack prove their worth by hitting back. - Piet Hein (via Tom White)
>>>>
>>>>
>>>>>________________________________
>>>>>From: Dhruba Borthakur <[EMAIL PROTECTED]>
>>>>>To: [EMAIL PROTECTED]
>>>>>Sent: Thursday, September 15, 2011 10:06 AM
>>>>>Subject: Re: Need help regarding HDFS-RAID
>>>>>
>>>>>
>>>>>
>>>>>We use HDFS RAID in a big way. Data older than 12 days are RAIDED using XOR encoding (effective replication of 2.5). Data older than a few months are raided using ReedSolomon (effective observed replication factor of 1.5). This is running on our 60 PB size cluster for about an year now.
>>>>>
>>>>>
>>>>>thanks
>>>>>dhruba
>>>>>
>>>>>
>>>>>
>>>>>On Thu, Sep 15, 2011 at 5:31 AM, Ajit Ratnaparkhi <[EMAIL PROTECTED]> wrote:
>>>>>
>>>>>Hi,
>>>>>>
>>>>>>
>>>>>>We were planning to use it for past data archival(instead of moving it to archival store).
>>>>>>Archiving it in HDFS gives advantage of making it easily available for processing whenever required.
>>>>>>
>>>>>>
>>>>>>Is there any archival solution in hadoop ecosystem?
>>>>>>
>>>>>>
>>>>>>thanks,
>>>>>>Ajit.
>>>>>>
>>>>>>
>>>>>>
>>>>>>On Thu, Sep 15, 2011 at 5:05 PM, Harsh J <[EMAIL PROTECTED]> wrote:
>>>>>>
>>>>>>Hey Ajit,
>>>>>>>
>>>>>>>HDFS-RAID was never part of the 0.20 release. It made its debut in the
>>>>>>>0.21 release [1]. I know that Facebook uses it (and also did develop
>>>>>>>it), but unsure of users beyond Facebook.
>>>>>>>
>>>>>>>While 0.21 overall is not entirely deemed as production-usable yet
>>>>>>>(and is in fact, possibly abandoned for efforts on 0.22+), you can
>>>>>>>give that release a whirl on a test cluster and see for yourself if
>>>>>>>your need beats the stability.
>>>>>>>
>>>>>>>Just curious though - why are you looking to use this specifically?
>>>>>>>
>>>>>>>[1] - http://svn.apache.org/repos/asf/hadoop/common/branches/branch-0.21/mapreduce/src/contrib/raid/
>>>>>>>
>>>>>>>
>>>>>>>On Thu, Sep 15, 2011 at 4:37 PM, Ajit Ratnaparkhi
>>>>>>><[EMAIL PROTECTED]> wrote: