Location via proxy:   [ UP ]  
[Report a bug]   [Manage cookies]                
SlideShare a Scribd company logo

Hadoop training in hyderabad-kellytechnologies

•Download as PPT, PDF•
•
K
Kelly Technologies

Hadoop Institutes : kelly technologies is the best Hadoop Training Institutes in Hyderabad. Providing Hadoop training by real time faculty in Hyderabad.

1 of 40
1 Hadoop File System Presented By
Reference 2 The Hadoop Distributed File System: Architecture and Design by www.kellytechno.com
Basic Features: HDFS 3 Highly fault-tolerant High throughput Suitable for applications with large data sets Streaming access to file system data Can be built out of commodity hardware www.kellytechno.com
Fault tolerance 4 Failure is the norm rather than exception A HDFS instance may consist of thousands of server machines, each storing part of the file system’s data. Since we have huge number of components and that each component has non-trivial probability of failure means that there is always some component that is non-functional. Detection of faults and quick, automatic recovery from them is a core architectural goal of HDFS. www.kellytechno.com
Data Characteristics 5 Streaming data access Applications need streaming access to data Batch processing rather than interactive user access. Large data sets and files: gigabytes to terabytes size High aggregate data bandwidth Scale to hundreds of nodes in a cluster Tens of millions of files in a single instance Write-once-read-many: a file once created, written and closed need not be changed – this assumption simplifies coherency A map-reduce application or web-crawler application fits perfectly with this model. www.kellytechno.com
Cat Bat Dog Other Words (size: TByte) map map map map split split split split combine combine combine reduce reduce reduce part0 part1 part2 MapReduce 6 www.kellytechno.com
Architecture 7 www.kellytechno.com
Namenode and Datanodes 8 Master/slave architecture HDFS cluster consists of a single Namenode, a master server that manages the file system namespace and regulates access to files by clients. There are a number of DataNodes usually one per node in a cluster. The DataNodes manage storage attached to the nodes that they run on. HDFS exposes a file system namespace and allows user data to be stored in files. A file is split into one or more blocks and set of blocks are stored in DataNodes. DataNodes: serves read, write requests, performs block creation, deletion, and replication upon instruction from Namenode. www.kellytechno.com
HDFS Architecture 9 Namenode B replication Rack1 Rack2 Client Blocks Datanodes Datanodes Client Write Read Metadata ops Metadata(Name, replicas..) (/home/foo/data,6. .. Block ops www.kellytechno.com
File system Namespace 10 Hierarchical file system with directories and files Create, remove, move, rename etc. Namenode maintains the file system Any meta information changes to the file system recorded by the Namenode. An application can specify the number of replicas of the file needed: replication factor of the file. This information is stored in the Namenode. www.kellytechno.com
Data Replication 11 HDFS is designed to store very large files across machines in a large cluster. Each file is a sequence of blocks. All blocks in the file except the last are of the same size. Blocks are replicated for fault tolerance. Block size and replicas are configurable per file. The Namenode receives a Heartbeat and a BlockReport from each DataNode in the cluster. BlockReport contains all the blocks on a Datanode. www.kellytechno.com
Replica Placement 12  The placement of the replicas is critical to HDFS reliability and performance.  Optimizing replica placement distinguishes HDFS from other distributed file systems.  Rack-aware replica placement:  Goal: improve reliability, availability and network bandwidth utilization  Research topic  Many racks, communication between racks are through switches.  Network bandwidth between machines on the same rack is greater than those in different racks.  Namenode determines the rack id for each DataNode.  Replicas are typically placed on unique racks  Simple but non-optimal  Writes are expensive  Replication factor is 3  Another research topic?  Replicas are placed: one on a node in a local rack, one on a different node in the local rack and one on a node in a different rack.  1/3 of the replica on a node, 2/3 on a rack and 1/3 distributed evenly across remaining racks. www.kellytechno.com
Replica Selection 13 Replica selection for READ operation: HDFS tries to minimize the bandwidth consumption and latency. If there is a replica on the Reader node then that is preferred. HDFS cluster may span multiple data centers: replica in the local data center is preferred over the remote one. www.kellytechno.com
Safemode Startup 14 On startup Namenode enters Safemode. Replication of data blocks do not occur in Safemode. Each DataNode checks in with Heartbeat and BlockReport. Namenode verifies that each block has acceptable number of replicas After a configurable percentage of safely replicated blocks check in with the Namenode, Namenode exits Safemode. It then makes the list of blocks that need to be replicated. Namenode then proceeds to replicate these blocks to other Datanodes. www.kellytechno.com
Filesystem Metadata 15 The HDFS namespace is stored by Namenode. Namenode uses a transaction log called the EditLog to record every change that occurs to the filesystem meta data. For example, creating a new file. Change replication factor of a file EditLog is stored in the Namenode’s local filesystem Entire filesystem namespace including mapping of blocks to files and file system properties is stored in a file FsImage. Stored in Namenode’s local filesystem. www.kellytechno.com
Namenode 16 Keeps image of entire file system namespace and file Blockmap in memory. 4GB of local RAM is sufficient to support the above data structures that represent the huge number of files and directories. When the Namenode starts up it gets the FsImage and Editlog from its local file system, update FsImage with EditLog information and then stores a copy of the FsImage on the filesytstem as a checkpoint. Periodic checkpointing is done. So that the system can recover back to the last checkpointed state in case of a crash. www.kellytechno.com
Datanode 17 A Datanode stores data in files in its local file system. Datanode has no knowledge about HDFS filesystem It stores each block of HDFS data in a separate file. Datanode does not create all files in the same directory. It uses heuristics to determine optimal number of files per directory and creates directories appropriately:  Research issue? When the filesystem starts up it generates a list of all HDFS blocks and send this report to Namenode: Blockreport. www.kellytechno.com
Protocol 18 www.kellytechno.com
The Communication Protocol 19 All HDFS communication protocols are layered on top of the TCP/IP protocol A client establishes a connection to a configurable TCP port on the Namenode machine. It talks ClientProtocol with the Namenode. The Datanodes talk to the Namenode using Datanode protocol. RPC abstraction wraps both ClientProtocol and Datanode protocol. Namenode is simply a server and never initiates a request; it only responds to RPC requests issued by DataNodes or clients. www.kellytechno.com
Robustness 20 www.kellytechno.com
Objectives 21 Primary objective of HDFS is to store data reliably in the presence of failures. Three common failures are: Namenode failure, Datanode failure and network partition. www.kellytechno.com
Objectives 22 Primary objective of HDFS is to store data reliably in the presence of failures. Three common failures are: Namenode failure, Datanode failure and network partition. www.kellytechno.com
Objectives 23 Primary objective of HDFS is to store data reliably in the presence of failures. Three common failures are: Namenode failure, Datanode failure and network partition. www.kellytechno.com
Objectives 24 Primary objective of HDFS is to store data reliably in the presence of failures. Three common failures are: Namenode failure, Datanode failure and network partition. www.kellytechno.com
DataNode failure and heartbeat 25 A network partition can cause a subset of Datanodes to lose connectivity with the Namenode. Namenode detects this condition by the absence of a Heartbeat message. Namenode marks Datanodes without Hearbeat and does not send any IO requests to them. Any data registered to the failed Datanode is not available to the HDFS. Also the death of a Datanode may cause replication factor of some of the blocks to fall below their specified value. www.kellytechno.com
Re-replication 26 The necessity for re-replication may arise due to: A Datanode may become unavailable, A replica may become corrupted, A hard disk on a Datanode may fail, or The replication factor on the block may be increased. www.kellytechno.com
Cluster Rebalancing 27 HDFS architecture is compatible with data rebalancing schemes. A scheme might move data from one Datanode to another if the free space on a Datanode falls below a certain threshold. In the event of a sudden high demand for a particular file, a scheme might dynamically create additional replicas and rebalance other data in the cluster. These types of data rebalancing are not yet implemented: research issue. www.kellytechno.com
Data Integrity 28 Consider a situation: a block of data fetched from Datanode arrives corrupted. This corruption may occur because of faults in a storage device, network faults, or buggy software. A HDFS client creates the checksum of every block of its file and stores it in hidden files in the HDFS namespace. When a clients retrieves the contents of file, it verifies that the corresponding checksums match. If does not match, the client can retrieve the block from a replica. www.kellytechno.com
Metadata Disk Failure 29 FsImage and EditLog are central data structures of HDFS. A corruption of these files can cause a HDFS instance to be non- functional. For this reason, a Namenode can be configured to maintain multiple copies of the FsImage and EditLog. Multiple copies of the FsImage and EditLog files are updated synchronously. Meta-data is not data-intensive. The Namenode could be single point failure: automatic failover is NOT supported! Another research topic. www.kellytechno.com
Data Organization 30 www.kellytechno.com
Data Blocks 31 HDFS support write-once-read-many with reads at streaming speeds. A typical block size is 64MB (or even 128 MB). A file is chopped into 64MB chunks and stored. www.kellytechno.com
Staging 32 A client request to create a file does not reach Namenode immediately. HDFS client caches the data into a temporary file. When the data reached a HDFS block size the client contacts the Namenode. Namenode inserts the filename into its hierarchy and allocates a data block for it. The Namenode responds to the client with the identity of the Datanode and the destination of the replicas (Datanodes) for the block. Then the client flushes it from its local memory. www.kellytechno.com
Staging (contd.) 33 The client sends a message that the file is closed. Namenode proceeds to commit the file for creation operation into the persistent store. If the Namenode dies before file is closed, the file is lost. This client side caching is required to avoid network congestion; also it has precedence is AFS (Andrew file system). www.kellytechno.com
Replication Pipelining 34 When the client receives response from Namenode, it flushes its block in small pieces (4K) to the first replica, that in turn copies it to the next replica and so on. Thus data is pipelined from Datanode to the next. www.kellytechno.com
API (Accessibility) 35 www.kellytechno.com
Application Programming Interface 36 HDFS provides Java API for application to use. Python access is also used in many applications. A C language wrapper for Java API is also available. A HTTP browser can be used to browse the files of a HDFS instance. www.kellytechno.com
FS Shell, Admin and Browser Interface 37 HDFS organizes its data in files and directories. It provides a command line interface called the FS shell that lets the user interact with data in the HDFS. The syntax of the commands is similar to bash and csh. Example: to create a directory /foodir /bin/hadoop dfs –mkdir /foodir There is also DFSAdmin interface available Browser interface is also available to view the namespace. www.kellytechno.com
Space Reclamation 38 When a file is deleted by a client, HDFS renames file to a file in be the /trash directory for a configurable amount of time. A client can request for an undelete in this allowed time. After the specified time the file is deleted and the space is reclaimed. When the replication factor is reduced, the Namenode selects excess replicas that can be deleted. Next heartbeat(?) transfers this information to the Datanode that clears the blocks for use. www.kellytechno.com
Summary 39 We discussed the features of the Hadoop File System, a peta- scale file system to handle big-data sets. What discussed: Architecture, Protocol, API, etc. Missing element: Implementation The Hadoop file system (internals) An implementation of an instance of the HDFS (for use by applications such as web crawlers). www.kellytechno.com
40 Presented By

More Related Content

Hadoop training in hyderabad-kellytechnologies

  • 2. Reference 2 The Hadoop Distributed File System: Architecture and Design by www.kellytechno.com
  • 3. Basic Features: HDFS 3 Highly fault-tolerant High throughput Suitable for applications with large data sets Streaming access to file system data Can be built out of commodity hardware www.kellytechno.com
  • 4. Fault tolerance 4 Failure is the norm rather than exception A HDFS instance may consist of thousands of server machines, each storing part of the file system’s data. Since we have huge number of components and that each component has non-trivial probability of failure means that there is always some component that is non-functional. Detection of faults and quick, automatic recovery from them is a core architectural goal of HDFS. www.kellytechno.com
  • 5. Data Characteristics 5 Streaming data access Applications need streaming access to data Batch processing rather than interactive user access. Large data sets and files: gigabytes to terabytes size High aggregate data bandwidth Scale to hundreds of nodes in a cluster Tens of millions of files in a single instance Write-once-read-many: a file once created, written and closed need not be changed – this assumption simplifies coherency A map-reduce application or web-crawler application fits perfectly with this model. www.kellytechno.com
  • 8. Namenode and Datanodes 8 Master/slave architecture HDFS cluster consists of a single Namenode, a master server that manages the file system namespace and regulates access to files by clients. There are a number of DataNodes usually one per node in a cluster. The DataNodes manage storage attached to the nodes that they run on. HDFS exposes a file system namespace and allows user data to be stored in files. A file is split into one or more blocks and set of blocks are stored in DataNodes. DataNodes: serves read, write requests, performs block creation, deletion, and replication upon instruction from Namenode. www.kellytechno.com
  • 9. HDFS Architecture 9 Namenode B replication Rack1 Rack2 Client Blocks Datanodes Datanodes Client Write Read Metadata ops Metadata(Name, replicas..) (/home/foo/data,6. .. Block ops www.kellytechno.com
  • 10. File system Namespace 10 Hierarchical file system with directories and files Create, remove, move, rename etc. Namenode maintains the file system Any meta information changes to the file system recorded by the Namenode. An application can specify the number of replicas of the file needed: replication factor of the file. This information is stored in the Namenode. www.kellytechno.com
  • 11. Data Replication 11 HDFS is designed to store very large files across machines in a large cluster. Each file is a sequence of blocks. All blocks in the file except the last are of the same size. Blocks are replicated for fault tolerance. Block size and replicas are configurable per file. The Namenode receives a Heartbeat and a BlockReport from each DataNode in the cluster. BlockReport contains all the blocks on a Datanode. www.kellytechno.com
  • 12. Replica Placement 12  The placement of the replicas is critical to HDFS reliability and performance.  Optimizing replica placement distinguishes HDFS from other distributed file systems.  Rack-aware replica placement:  Goal: improve reliability, availability and network bandwidth utilization  Research topic  Many racks, communication between racks are through switches.  Network bandwidth between machines on the same rack is greater than those in different racks.  Namenode determines the rack id for each DataNode.  Replicas are typically placed on unique racks  Simple but non-optimal  Writes are expensive  Replication factor is 3  Another research topic?  Replicas are placed: one on a node in a local rack, one on a different node in the local rack and one on a node in a different rack.  1/3 of the replica on a node, 2/3 on a rack and 1/3 distributed evenly across remaining racks. www.kellytechno.com
  • 13. Replica Selection 13 Replica selection for READ operation: HDFS tries to minimize the bandwidth consumption and latency. If there is a replica on the Reader node then that is preferred. HDFS cluster may span multiple data centers: replica in the local data center is preferred over the remote one. www.kellytechno.com
  • 14. Safemode Startup 14 On startup Namenode enters Safemode. Replication of data blocks do not occur in Safemode. Each DataNode checks in with Heartbeat and BlockReport. Namenode verifies that each block has acceptable number of replicas After a configurable percentage of safely replicated blocks check in with the Namenode, Namenode exits Safemode. It then makes the list of blocks that need to be replicated. Namenode then proceeds to replicate these blocks to other Datanodes. www.kellytechno.com
  • 15. Filesystem Metadata 15 The HDFS namespace is stored by Namenode. Namenode uses a transaction log called the EditLog to record every change that occurs to the filesystem meta data. For example, creating a new file. Change replication factor of a file EditLog is stored in the Namenode’s local filesystem Entire filesystem namespace including mapping of blocks to files and file system properties is stored in a file FsImage. Stored in Namenode’s local filesystem. www.kellytechno.com
  • 16. Namenode 16 Keeps image of entire file system namespace and file Blockmap in memory. 4GB of local RAM is sufficient to support the above data structures that represent the huge number of files and directories. When the Namenode starts up it gets the FsImage and Editlog from its local file system, update FsImage with EditLog information and then stores a copy of the FsImage on the filesytstem as a checkpoint. Periodic checkpointing is done. So that the system can recover back to the last checkpointed state in case of a crash. www.kellytechno.com
  • 17. Datanode 17 A Datanode stores data in files in its local file system. Datanode has no knowledge about HDFS filesystem It stores each block of HDFS data in a separate file. Datanode does not create all files in the same directory. It uses heuristics to determine optimal number of files per directory and creates directories appropriately:  Research issue? When the filesystem starts up it generates a list of all HDFS blocks and send this report to Namenode: Blockreport. www.kellytechno.com
  • 19. The Communication Protocol 19 All HDFS communication protocols are layered on top of the TCP/IP protocol A client establishes a connection to a configurable TCP port on the Namenode machine. It talks ClientProtocol with the Namenode. The Datanodes talk to the Namenode using Datanode protocol. RPC abstraction wraps both ClientProtocol and Datanode protocol. Namenode is simply a server and never initiates a request; it only responds to RPC requests issued by DataNodes or clients. www.kellytechno.com
  • 21. Objectives 21 Primary objective of HDFS is to store data reliably in the presence of failures. Three common failures are: Namenode failure, Datanode failure and network partition. www.kellytechno.com
  • 22. Objectives 22 Primary objective of HDFS is to store data reliably in the presence of failures. Three common failures are: Namenode failure, Datanode failure and network partition. www.kellytechno.com
  • 23. Objectives 23 Primary objective of HDFS is to store data reliably in the presence of failures. Three common failures are: Namenode failure, Datanode failure and network partition. www.kellytechno.com
  • 24. Objectives 24 Primary objective of HDFS is to store data reliably in the presence of failures. Three common failures are: Namenode failure, Datanode failure and network partition. www.kellytechno.com
  • 25. DataNode failure and heartbeat 25 A network partition can cause a subset of Datanodes to lose connectivity with the Namenode. Namenode detects this condition by the absence of a Heartbeat message. Namenode marks Datanodes without Hearbeat and does not send any IO requests to them. Any data registered to the failed Datanode is not available to the HDFS. Also the death of a Datanode may cause replication factor of some of the blocks to fall below their specified value. www.kellytechno.com
  • 26. Re-replication 26 The necessity for re-replication may arise due to: A Datanode may become unavailable, A replica may become corrupted, A hard disk on a Datanode may fail, or The replication factor on the block may be increased. www.kellytechno.com
  • 27. Cluster Rebalancing 27 HDFS architecture is compatible with data rebalancing schemes. A scheme might move data from one Datanode to another if the free space on a Datanode falls below a certain threshold. In the event of a sudden high demand for a particular file, a scheme might dynamically create additional replicas and rebalance other data in the cluster. These types of data rebalancing are not yet implemented: research issue. www.kellytechno.com
  • 28. Data Integrity 28 Consider a situation: a block of data fetched from Datanode arrives corrupted. This corruption may occur because of faults in a storage device, network faults, or buggy software. A HDFS client creates the checksum of every block of its file and stores it in hidden files in the HDFS namespace. When a clients retrieves the contents of file, it verifies that the corresponding checksums match. If does not match, the client can retrieve the block from a replica. www.kellytechno.com
  • 29. Metadata Disk Failure 29 FsImage and EditLog are central data structures of HDFS. A corruption of these files can cause a HDFS instance to be non- functional. For this reason, a Namenode can be configured to maintain multiple copies of the FsImage and EditLog. Multiple copies of the FsImage and EditLog files are updated synchronously. Meta-data is not data-intensive. The Namenode could be single point failure: automatic failover is NOT supported! Another research topic. www.kellytechno.com
  • 31. Data Blocks 31 HDFS support write-once-read-many with reads at streaming speeds. A typical block size is 64MB (or even 128 MB). A file is chopped into 64MB chunks and stored. www.kellytechno.com
  • 32. Staging 32 A client request to create a file does not reach Namenode immediately. HDFS client caches the data into a temporary file. When the data reached a HDFS block size the client contacts the Namenode. Namenode inserts the filename into its hierarchy and allocates a data block for it. The Namenode responds to the client with the identity of the Datanode and the destination of the replicas (Datanodes) for the block. Then the client flushes it from its local memory. www.kellytechno.com
  • 33. Staging (contd.) 33 The client sends a message that the file is closed. Namenode proceeds to commit the file for creation operation into the persistent store. If the Namenode dies before file is closed, the file is lost. This client side caching is required to avoid network congestion; also it has precedence is AFS (Andrew file system). www.kellytechno.com
  • 34. Replication Pipelining 34 When the client receives response from Namenode, it flushes its block in small pieces (4K) to the first replica, that in turn copies it to the next replica and so on. Thus data is pipelined from Datanode to the next. www.kellytechno.com
  • 36. Application Programming Interface 36 HDFS provides Java API for application to use. Python access is also used in many applications. A C language wrapper for Java API is also available. A HTTP browser can be used to browse the files of a HDFS instance. www.kellytechno.com
  • 37. FS Shell, Admin and Browser Interface 37 HDFS organizes its data in files and directories. It provides a command line interface called the FS shell that lets the user interact with data in the HDFS. The syntax of the commands is similar to bash and csh. Example: to create a directory /foodir /bin/hadoop dfs –mkdir /foodir There is also DFSAdmin interface available Browser interface is also available to view the namespace. www.kellytechno.com
  • 38. Space Reclamation 38 When a file is deleted by a client, HDFS renames file to a file in be the /trash directory for a configurable amount of time. A client can request for an undelete in this allowed time. After the specified time the file is deleted and the space is reclaimed. When the replication factor is reduced, the Namenode selects excess replicas that can be deleted. Next heartbeat(?) transfers this information to the Datanode that clears the blocks for use. www.kellytechno.com
  • 39. Summary 39 We discussed the features of the Hadoop File System, a peta- scale file system to handle big-data sets. What discussed: Architecture, Protocol, API, etc. Missing element: Implementation The Hadoop file system (internals) An implementation of an instance of the HDFS (for use by applications such as web crawlers). www.kellytechno.com