trafodion-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Sean Broeder <sean.broe...@esgyn.com>
Subject RE: A few questions
Date Fri, 23 Dec 2016 18:46:33 GMT
Rinka,
As Rohit mentioned HBase is not exactly pluggable into Trafodion.  I think more specifically
you seem to want to know if the storage engine in interchangeable.  It could be done, and
in fact HBase is not the first storage engine that that our SQL and Transaction engines have
utilized, but the current incarnation makes heavy use of HBase coprocessors and this would
be rather cumbersome to remove or duplicate in a replacement engine.  Still, Trafodion is
open source, so with additional contributors come additional possibilities and we would welcome
that.

Cheers,
Sean

-----Original Message-----
From: Rohit Jain [mailto:rohit.jain@esgyn.com] 
Sent: Friday, December 23, 2016 10:27 AM
To: dev@trafodion.incubator.apache.org
Subject: RE: A few questions

Rinka,

Here is a response to your questions:

1.  Is HBase pluggable?
For a query engine like Trafodion to get the best out of a storage engine like HBase, very
deep integration is required.  We can certainly provide a list of the deep integration we
have done with HBase.  But suffice is to say that to support another storage engine requires
a few months of development and performance tuning effort.  So while there is nothing preventing
Trafodion from integrating with other storage engines, a fair amount of effort is required
to do so well.  In that sense, another storage engine is not "pluggable".  Esgyn has a supported
and Enterprise database engine called EsgynDB, that is powered by Trafodion, which integrates
with Apache ORC as well to support BI and Analytics workloads for example.

 2.  Is Trafodion production ready?
Trafodion has been production ready for OLTP since R1.0 which was released very early 2015.
 So yes, it has been production ready for two years now and is in production at a number of
customer sites.

3.  OLTP workloads supported?
Actually, while there are no in-place updates supported by HBase, it is an MVCC model where
an update translates to a new version of the row being inserted.  This is a model that is
followed by a number of database implementations and not just HBase.  You can choose the number
of versions of a row you want to keep, thereby providing temporal and change data capture
capabilities as well.  With compaction, old versions are deleted and the new versions are
integrated into the main HFile.  With this MVCC model Trafodion supports full OLTP capabilities
with very impressive OLTP performance proven by internal benchmarks we run at Esgyn based
on TPC-C.

Please feel free to contact us for more detailed information, presentations, and architectural
capabilities between Trafodion and HBase.

Rohit

-----Original Message-----
From: Rinka Singh [mailto:rinka.singh@gmail.com] 
Sent: Friday, December 23, 2016 2:34 PM
To: dev@trafodion.incubator.apache.org
Subject: A few questions

Hi,
we are an open source startup and are considering using Trafordion for 
various deployments as well as contributing to it.  We have a customer 
where we would like to propose Trafordion as we see a pretty good fit.

We had the following questions for you:
* Is HBase pluggable.  How easy would it be to plug another storage 
engine instead of HBase?  What is involved?
* Is Trafordion production ready.  We have a mid-sized customer who 
would like to use it for OLTP transactions.  Can we put it into their 
production environment?
* What kind of OLTP workloads is HBase suitable for if there is no 
in-place updates in HDFS?

Looking forward to a quick reply,

Thanks in advance,

Rinka.

-- 
Rinka Singh
CoFounder, Melt Iron             Accelerating the Enterprise
............................................................

Mime
View raw message