[
https://issues.apache.org/jira/browse/HADOOP-692?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#action_12463383
]
Hairong Kuang commented on HADOOP-692:
--------------------------------------
The attached patch includes:
1. data structures for describing cluster network topology
a. org.apache.hadoop.net.NetworkTopology: class for manipulating the
tree-like cluster topology.
. InnerNode: represents racks/datacenters
. Leaves: represents datanodes
b. org.apache.hadoop.net.Node: class for abstracting a node in
NetworkTopology
. name: name of the node
. networklocation: indicating which datacenter and rack that this node is
located.
c. org.apchae.hadoop.net.NodeBase: a base implementation of Node
2. code for implementing replica placement strategy
a org.apache.hadoop.dfs.FSNamesystem.Replicator
. chooseTarget(numOfReplicas, writer, excludedNodes, blockSize): choose
numOfReplcas targets
. chooseTarget(numOfReplicas, writer, choosenNodes, excludedNodes, block):
repreplicate numOfReplicas targets
Targets are returned forming a shortest path that starts from writer and
traversing all targets
3. protocol changes
a. org.apache.hadoop.dfs.ClientProtocol
. open is enhanced by passing the clientMachine name. In the returned
blocks, containing data nodes are sorted by its distance to the clientMachine
b. org.apache.hadoop.dfs.DatanodeProtocol
. register is enhanced by passing the data node network location.
NetworkTopology is built while registering data nodes.
4. data structures changes to reflecting data node network location information
a. org.apache.hadoop.dfs.Datanode
b. org.apache.hadoop.dfs.DatanodeInfo
5. various test cases
a. org.apache.hadoop.net.TestNetworkTopolgy: test NetworkTopology
b. org.apache.hadoop.dfs.TestReplicationPolocy: test replicas placement
c. org.apache.hadoop.dfs.TestRackAwareReplication: test replicas
placement/datanode registration by bringing up a MiniDFSCluster
d. org.apache.hadoop.dfs.MiniDFSCluster: enhance it to start datanodes on
different network locations
Please review! I appreciate your feedback.
> Rack-aware Replica Placement
> ----------------------------
>
> Key: HADOOP-692
> URL: https://issues.apache.org/jira/browse/HADOOP-692
> Project: Hadoop
> Issue Type: Improvement
> Components: dfs
> Affects Versions: 0.8.0
> Reporter: Hairong Kuang
> Assigned To: Hairong Kuang
> Attachments: rack.patch, Rack_aware_HDFS_proposal.pdf
>
>
> This issue assumes that HDFS runs on a cluster of computers that spread
> across many racks. Communication between two nodes on different racks needs
> to go through switches. Bandwidth in/out of a rack may be less than the total
> bandwidth of machines in the rack. The purpose of rack-aware replica
> placement is to improve data reliability, availability, and network bandwidth
> utilization. The basic idea is that each data node determines to which rack
> it belongs at the startup time and notifies the name node of the rack id upon
> registration. The name node maintains a rackid-to-datanode map and tries to
> place replicas across racks.
--
This message is automatically generated by JIRA.
-
If you think it was sent incorrectly contact one of the administrators:
https://issues.apache.org/jira/secure/Administrators.jspa
-
For more information on JIRA, see: http://www.atlassian.com/software/jira
