I am thinking of storing medium sized objects (~1M) using HBase. The
advantage of using HBase rather than HBase (storing pointers) + HDFS, in
my mind, is:
data locality. When I want to run analytics, I will access these objects
using HBase scan, and HBase stores KVs in a sequential manner. If I use
HDFS, there is no guarantee that row 1 and row 2's files are adjacent to
each other.
store small files in HDFS is not efficient. Facebook's Haystack sort of
stitch small files together, while HBase achieve the same effect.
Any disadvantage I missed? I am also thinking of larger block size given
the object size.
Thanks,
Wei
From: Upendra Yadav <upendra1...@gmail.com>
To: user@hbase.apache.org,
Date: 02/25/2014 03:31 PM
Subject: Re: Is HBase is feasible for storing 4-5 MB of data as
cell value
Me too realize same what you suggest...: (Keep them in a separate files in
HDFS and store in HBase only references)
will try several attachments into a single file...
And Thanks a lot...
On Wed, Feb 26, 2014 at 1:45 AM, Vladimir Rodionov
<vrodio...@carrieriq.com>wrote:
> Usually, it is not advisable to store such a large values in HBase (to
> avoid excessive IO during compaction).
> Keep them in a separate files in HDFS and store in HBase only
references.
> To overcome inherent max file number limitation of NN
> you can bulk several values into a single file (you will need separate
> process -M/R job to garbage collect expired or deleted items).
>
> Best regards,
> Vladimir Rodionov
> Principal Platform Engineer
> Carrier IQ, www.carrieriq.com
> e-mail: vrodio...@carrieriq.com
>
> ________________________________________
> From: Ted Yu [yuzhih...@gmail.com]
> Sent: Tuesday, February 25, 2014 12:02 PM
> To: user@hbase.apache.org
> Subject: Re: Is HBase is feasible for storing 4-5 MB of data as cell
value
>
> Minor:
> Value 0 also means no cap - see HTable#validatePut()
>
> if (maxKeyValueSize > 0) {
>
> ...
>
> if (kv.getLength() > maxKeyValueSize) {
>
> throw new IllegalArgumentException("KeyValue size too
large");
>
> }
>
>
> On Tue, Feb 25, 2014 at 11:52 AM, Ameya Kanitkar <am...@groupon.com>
> wrote:
>
> > The only other thing I'd add is, by default HBase caps size of the
data
> per
> > column at 10 MB (I think). You can change that by changing this
setting:
> >
> > hbase.client.keyvalue.maxsize
> > in hbase-site.xml
> >
> > -1 means no cap. You can put other numbers for appropriate cap for
your
> use
> > case.
> >
> > Ameya
> >
> >
> > On Tue, Feb 25, 2014 at 12:12 AM, shashwat shriparv <
> > dwivedishash...@gmail.com> wrote:
> >
> > > Yes for sure you can use hbase for this, you can have
> > > 1. different fields of mail in different column of a column family
and
> > > attachment as a binary array also in a column.
> > > 2. you can keep whole message in columns in hbase and the
attachments
> are
> > > large enoug on the hdfs and some reference to it in hbase table.
> > > 3. schema you can decide, you can have a matrix how you store values
to
> > > that you can decide.
> > >
> > >
> > > *Warm Regards_**∞_*
> > > * Shashwat Shriparv*
> > > [image: http://www.linkedin.com/pub/shashwat-shriparv/19/214/2a9]<
> > > http://www.linkedin.com/pub/shashwat-shriparv/19/214/2a9>[image:
> > > https://twitter.com/shriparv] <https://twitter.com/shriparv>[image:
> > > https://www.facebook.com/shriparv] <
https://www.facebook.com/shriparv
> > > >[image:
> > > http://google.com/+ShashwatShriparv]
> > > <http://google.com/+ShashwatShriparv>[image:
> > > http://www.youtube.com/user/sShriparv/videos]<
> > > http://www.youtube.com/user/sShriparv/videos>[image:
> > > http://profile.yahoo.com/SWXSTW3DVSDTF2HHSRM47AV6DI/] <
> > shrip...@yahoo.com>
> > >
> > >
> > >
> > > On Tue, Feb 25, 2014 at 12:55 PM, Upendra Yadav
<upendra1...@gmail.com
> > > >wrote:
> > >
> > > > I have to use hbase and have mix type of data
> > > >
> > > > Some of them have size 1-4K(Mail- Header....) and others
> > > > >5MB(Attachments...)
> > > >
> > > > And also we need only random access: any data
> > > >
> > > > Is HBase is feasible for storing this type of data
> > > >
> > > > What will be my schema design -
> > > > will have to go with 2 different Table -> 1st one for 1-4K and
2nd
> for
> > > big
> > > > file
> > > > (because of memstore flush will flush other CF, and huge random
> access)
> > > >
> > > > Or there is other way:;
> > > >
> > > > Thanks
> > > >
> > >
> >
>
> Confidentiality Notice: The information contained in this message,
> including any attachments hereto, may be confidential and is intended to
be
> read only by the individual or entity to whom this message is addressed.
If
> the reader of this message is not the intended recipient or an agent or
> designee of the intended recipient, please note that any review, use,
> disclosure or distribution of this message or its attachments, in any
form,
> is strictly prohibited. If you have received this message in error,
please
> immediately notify the sender and/or notificati...@carrieriq.com and
> delete or destroy any copy of this message and its attachments.
>
