Hadoop training institutes in hyderabad

Page 1

Presented By Kelly Technologies www.kellytechno.com


Agenda  What is distributed computing  Why distributed computing  Common Architecture  Best Practice  Case study  Condor  Hadoop – HDFS and map reduce

www.kellytechno.com


What is Distributed Computing/System?  Distributed computing  A field of computing science that

studies distributed system.  The use of distributed systems to solve computational problems.  Distributed system  Wikipedia 

There are several autonomous computational entities, each of which has its own local memory. The entities communicate with each other by message passing.

 Operating System Concept 

The processors communicate with one another through various communication lines, such as high-speed buses or telephone lines. Each processor has its own local memory.

www.kellytechno.com


What is Distributed Computing/System?  Distributed program  A computing program that runs in a distributed system  Distributed programming  The process of writing distributed program

www.kellytechno.com


What is Distributed Computing/System?  Common properties  Fault tolerance  

When one or some nodes fails, the whole system can still work fine except performance. Need to check the status of each node

 Each node play partial role  Each computer has only a limited, incomplete view of the system. Each computer may know only one part of the input.  Resource sharing  Each user can share the computing power and storage resource in the system with other users  Load Sharing  Dispatching several tasks to each nodes can help share loading to the whole system.  Easy to expand  We expect to use few time when adding nodes. Hope to spend no time if possible.

www.kellytechno.com


Why Distributed Computing?  The nature of application  Performance  Computing intensive 

The task could consume a lot of time on computing. For example, π

 Data intensive 

The task that deals with a lot mount or large size of files. For example, Facebook, LHC(Large Hadron Collider).

 Robustness  No SPOF (Single Point Of Failure)  Other nodes can execute the same task executed on failed node.

www.kellytechno.com


Common Architectures  Communicate and coordinate works among concurrent

processes

 Processes communicate by sending/receiving messages  Synchronous/Asynchronous

www.kellytechno.com


Common Architectures  Master/Slave architecture  Master/slave is a model of communication where one device or process has unidirectional control over one or more other devices 

Database replication  Source database can be treated as a master and the destination database can treated as a slave. Client-server  web browsers and web servers

www.kellytechno.com


Common Architectures  Data-centric architecture  Using a standard, general-purpose relational database

management system  customized in-memory or file-based data structures and access method  Using dynamic, table-driven logic in  logic embodied in previously compiled programs  Stored procedures  logic running in middle-tier application servers  Shared databases as the basis for communicating between parallel processes  direct inter-process communication via message passing function

www.kellytechno.com


Best Practice

 Data Intensive or Computing Intensive  Data size and the amount of data   

The attribute of data you consume Computing intensive  We can move data to the nodes where we can execute jobs Data Intensive  We can separate/replicate data to difference nodes, then we can execute our tasks on these nodes  Reduce data replication when executing tasks

 Master nodes need to know data location  No data loss when incidents happen  SAN (Storage Area Network)  Data replication on different nodes  Synchronization  When splitting tasks to different nodes, how can we make sure these tasks are synchronized?

www.kellytechno.com


Best Practice  Robustness  Still safe when one or partial nodes fail  Need to recover when failed nodes are online. No further or few action is needed 

Condor – restart daemon

 Failure detection 

When any nodes fails, master nodes can detect this situation.  Eg: Heartbeat detection

 App/Users don’t need to know if any partial failure

happens. 

Restart tasks on other nodes for users

www.kellytechno.com


Best Practice  Network issue  Bandwidth 

Need to think of bandwidth when copying files from one node to other nodes if we would like to execute the task on the nodes if no data in these nodes.

 Scalability  Easy to expand 

Hadoop – configuration modification and start daemon

 Optimization  What can we do if the performance of some nodes is not good? 

Monitoring the performance of each node  According to any information exchange like heartbeat or log Resume the same task on another nodes

www.kellytechno.com


Best Practice  App/User  shouldn’t know how to communicate between nodes  User mobility – user can access the system from some point or anywhere  

Grid – UI (User interface) Condor – submit machine

www.kellytechno.com


Case study - Condor  Condor  Computing intensive jobs  Queuing policy 

Match task and computing nodes

 Resource Classification 

Each resource can advertise its attributes and master can classify according to this

www.kellytechno.com


Case study - Condor

www.kellytechno.com


Case study - Condor  Role  Central Manger 

The collector of information, and the negotiator between resources and resource requests

 Execution machine 

Responsible for executing condor tasks

 Submit machine 

Responsible for submitting condor tasks

 Checkpoint servers 

Responsible for storing all checkpoint files for the tasks

www.kellytechno.com


Case study - Condor  Robustness  One execution machine fails 

We can execute the same task on other nodes.

 Recovery 

Only need to restart the daemon when the failed nodes are online

www.kellytechno.com


Case study - Condor  Resource sharing  Each condor user can share computing power with other condor users.  Synchronization  Users need to take care by themselves 

Users can execute MPI job in a condor pool but need to think of the issues of synchronization and Deadlock.

 Failure detection  Central manager can know when nodes fails 

Based on update notification sent by nodes

 Scalability  Only execute few commands when new nodes are online.

www.kellytechno.com


Case study - Hadoop  HDFS  Namenode:  

manages the file system namespace and regulates access to files by clients. determines the mapping of blocks to DataNodes.

 Data Node :  manage storage attached to the nodes that they run on  save CRC codes  send heartbeat to namenode.  Each data is split as a chunk and each chuck is stored on some data nodes.  Secondary Namenode 

responsible for merging fsImage and EditLog

www.kellytechno.com


Case study - Hadoop

www.kellytechno.com


Case study - Hadoop  Map-reduce Framework  JobTracker  

Responsible for dispatch job to each tasktracker Job management like removing and scheduling.

 TaskTracker 

Responsible for executing job. Usually tasktracker launch another JVM to execute the job.

www.kellytechno.com


Case study - Hadoop

www.kellytechno.com

From Hadoop - The Definitive Guide


Case study - Hadoop  Data replication  Data are replicated to different nodes  

Reduce the possibility of data loss Data locality. Job will be sent to the node where data are.

 Robustness  One datanode fails 

We can get data from other nodes.

 One tasktracker failed 

We can start the same task on different node

 Recovery 

Only need to restart the daemon when the failed nodes are online

www.kellytechno.com


Case study - Hadoop  Resource sharing  Each hadoop user can share computing power and storage space with other hadoop users.  Synchronization  No synchronization

 Failure detection  Namenode/Jobtracker can know when datanode/tasktracker fails 

Based on heartbeat

www.kellytechno.com


Case study - Hadoop  Scalability  Only execute few commands when new nodes are online.  Optimization  A speculative task is launched only when a task takes

too much time on one node. 

The slower task will be killed when the other one has been finished

www.kellytechno.com


www.kellytechno.com


Turn static files into dynamic content formats.

Create a flipbook
Issuu converts static files into: digital portfolios, online yearbooks, online catalogs, digital photo albums and more. Sign up and create your flipbook.