The Hadoop Distributed File System

Konstantin Shvachko, Hairong Kuang, Sanjay Radia, Robert Chansler Yahoo! Sunnyvale, California USA

Presented by Ying Yang 9/24/2012

Outline
Introduction Architecture
NameNode, DataNodes, HDFS Client, CheckpointNode, BackupNode, Snapshots

File I/O Operations and Replica Management

File Read and Write, Block Placement, Replication management, Balancer,

Practice at YAHoo! FUTURE WORK

Introduction
HDFS
The Hadoop Distributed File System (HDFS) is the file system component of Hadoop. It is designed to store very large data sets (1) reliably, and to stream those data sets (2) at high bandwidth to user applications. These are achieved by replicating file content on multiple machines(DataNodes).

Outline
Introduction Architecture
NameNode, DataNodes, HDFS Client, CheckpointNode, BackupNode, Snapshots

File I/O Operations and Replica Management

File Read and Write, Block Placement, Replication management, Balancer,

Practice at YAHoo! FUTURE WORK

Architecture
HDFS is a block-structured file system: Files broken into blocks of 128MB (per-file configurable). A file can be made of several blocks, and they are stored across a cluster of one or more machines with data storage capacity. Each block of a file is replicated across a number of machines, To prevent loss of data.

Architecture

Architecture
NameNode and DataNodes HDFS stores file system metadata and application data separately. Metadata refers to file metadata(attributes such as permissions, modification, access times, namespace and disk space quotas. )called inodes+list of blocks belong to the file. HDFS stores metadata on a dedicated server, called the NameNode.(Master) Application data are stored on other servers called DataNodes.(Slaves) All servers are fully connected and communicate with each other using TCP-based protocols.(RPC)

Architecture
Single Namenode:
Maintain the namespace tree(a hierarchy of files and directories) operations like opening, closing, and renaming files and directories. Determine the mapping of file blocks to DataNodes (the physical location of file data). File metadata (i.e. inode) . Authorization and authentication. Collect block reports from Datanodes on block locations. Replicate missing blocks.
HDFS keeps the entire namespace in RAM, allowing fast access to the metadata.

Architecture
DataNodes:
The DataNodes are responsible for serving read and write requests from the file systems clients. The DataNodes also perform block creation, deletion, and replication upon instruction from the NameNode.
Data nodes periodically send block reports to Namenode.

Architecture

Architecture

Architecture

Architecture

Architecture

Architecture
NameNode and DataNode communication: Heartbeats.

DataNodes send heartbeats to the NameNode to confirm that the DataNode is operating and the block replicas it hosts are available.

Architecture

Architecture
Blockreports:
A DataNode identifies block replicas in its possession to the NameNode by sending a block report. A block report contains the block id, the generation stamp and the length for each block replica the server hosts. Blockreports provide the NameNode with an up-to-date view of where block replicas are located on the cluster and nameNode constructs and maintains latest metadata from blockreports.

Architecture

Architecture
failure recovery
The NameNode does not directly call DataNodes. It uses replies to heartbeats to send instructions to the DataNodes. The instructions include commands to: replicate blocks to other nodes: DataNode died. copy data to local. remove local block replicas; re-register or to shut down the node;

Architecture

Architecture
failure recovery So when dataNode died, NameNode will notice and instruct other dataNode to replicate data to new dataNode. What if NameNode died?

Architecture
failure recovery Keep journal (the modification log of metadata). Checkpoint: The persistent record of the metadata stored in the local hosts native files system. For example: During restart, the NameNode initializes the namespace image from the checkpoint, and then replays changes from the journal until the image is up-to-date with the last state of the file system.

Architecture
failure recovery CheckpointNode and BackupNode--two other roles of NameNode

CheckpointNode: When journal becomes too long, checkpointNode combines the existing checkpoint and journal to create a new checkpoint and an empty journal.

Architecture
failure recovery CheckpointNode and BackupNode--two other roles of NameNode BackupNode: A read-only NameNode it maintains an in-memory, up-to-date image of the file system namespace that is always synchronized with the state of the NameNode. If the NameNode fails, the BackupNodes image in memory and the checkpoint on disk is a record of the latest namespace state.

Architecture
failure recovery Upgrades, File System Snapshots The purpose of creating snapshots in HDFS is to minimize potential damage to the data stored in the system during upgrades. During software upgrades the possibility of corrupting the system due to software bugs or human mistakes increases. The snapshot mechanism lets administrators persistently save the current state of the file system(both data and metadata), so that if the upgrade results in data loss or corruption, it is possible to rollback the upgrade and return HDFS to the namespace and storage state as they were at the time of the snapshot.

Outline
Introduction Architecture
NameNode, DataNodes, HDFS Client, CheckpointNode, BackupNode, Snapshots

File I/O Operations and Replica Management

File Read and Write, Block Placement, Replication management, Balancer,

Practice at YAHoo! FUTURE WORK

File I/O Operations and Replica Management

File I/O Operations and Replica Management

Hadoop has the concept of Rack Awareness.

File I/O Operations and Replica Management

Hadoop has the concept of Rack Awareness. The default HDFS replica placement policy can be summarized as follows: 1. No Datanode contains more than one replica of any block. 2. No rack contains more than two replicas of the same block, provided there are sufficient racks on the cluster.

File I/O Operations and Replica Management

File I/O Operations and Replica Management

OK. Write to dataNode 1.
dsHey,DN1, duplicate Block A to DN5 and DN6.

File I/O Operations and Replica Management

File I/O Operations and Replica Management

File I/O Operations and Replica Management

File I/O Operations and Replica Management

File I/O Operations and Replica Management

Balancer

File I/O Operations and Replica Management

Balancer

Outline
Introduction Architecture
NameNode, DataNodes, HDFS Client, CheckpointNode, BackupNode, Snapshots

File I/O Operations and Replica Management

File Read and Write, Block Placement, Replication management, Balancer,

Practice at YAHoo! FUTURE WORK

Practice at YAHoo!
HDFS clusters at Yahoo! include about 3500 nodes A typical cluster node has:
2 quad core Xeon processors @ 2.5ghz Red Hat Enterprise Linux Server Release 5.1 Sun Java JDK 1.6.0_13-b03 4 directly attached SATA drives (one terabyte each) 16G RAM 1-gigabit Ethernet

Practice at YAHoo!
70 percent of the disk space is allocated to HDFS. The remainder is reserved for the operating system (Red Hat Linux), logs, and space to spill the output of map tasks. (MapReduce intermediate data are not stored in HDFS.) For each cluster, the NameNode and the BackupNode hosts are specially provisioned with up to 64GB RAM; application tasks are never assigned to those hosts. In total, a cluster of 3500 nodes has 9.8 PB of storage available as blocks that are replicated three times yielding a net 3.3 PB of storage for user applications. As a convenient approximation, one thousand nodes represent one PB of application storage.

Practice at YAHoo!
Durability of Data uncorrelated node failures Replication of data three times is a robust guard against loss of data due to uncorrelated node failures. correlated node failures, the failure of a rack or core switch. HDFS can tolerate losing a rack switch (each block has a replica on some other rack). loss of electrical power to the cluster a large cluster will lose a handful of blocks during a power-on restart.

Practice at YAHoo!
Benchmarks

Practice at YAHoo!
Benchmarks

NameNode Throughput benchmark

FUTURE WORK
Automated failover

plan: Zookeeper, Yahoos distributed consensus technology to build an automated failover solution Scalability of the NameNode Solution: Our near-term solution to scalability is to allow multiple namespaces (and NameNodes) to share the physical storage within a cluster. Drawbacks: The main drawback of multiple independent namespaces is the cost of managing them.