hadoop-common-commits mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Apache Wiki <wikidi...@apache.org>
Subject [Hadoop Wiki] Trivial Update of "Hamburg" by HyunsikChoi
Date Mon, 14 Sep 2009 04:44:29 GMT
Dear Wiki user,

You have subscribed to a wiki page or wiki category on "Hadoop Wiki" for change notification.

The following page has been changed by HyunsikChoi:
http://wiki.apache.org/hadoop/Hamburg

------------------------------------------------------------------------------
  
  == Goal ==
   * Support graph traverse
-  * Support a simple programming interface dealing graph data
+  * Support a simple programming interface familiar with graph features.
   * Follow the scalability concept of shared-nothing architecture
   * Fault-Tolerant Implementation
  
  == Hamburg ==
- Hambrug is an alternative to MR programming model. It consists of two parts, each of which
is related to locality-preserving storage method for graph in terms of connectivity and computations
with traverse interface on graphs respectively.
+ Hambrug is an alternative to MR programming model. It consists of two parts, each of which
is related to locality-preserving storing method for graph in terms of connectivity and computations
with traverse interface on graphs respectively.
  
- The main purpose of Locality-preserving storage methods for graph is to store vertices close
to one another into the same HDFS block. The computation part with this may reduce considerable
communication cost and the number of bulk sync step. It will be a kind of preprocessed step
and be implemented in MR.
+ The main purpose of the locality-preserving storing method for graph is to store vertices
close to one another into the same HDFS block. The computation part with this storing method
may reduce considerable communication cost and the number of bulk sync step. It will be a
kind of pre-process step and be implemented in MR.
  
  The computation part is based on bulk synchronization parallel (BSP) model. Like MR, Hamburg
will take advantages from shared-nothing architecture (SN), so I expect that it will also
show scalability without almost degradation of performance as the number of participant nodes
increases. In addition, we will provide a set of easy APIs familiar with graph features and
similar to MR.
  

Mime
View raw message