HBase and HDFS: Understanding FileSystem Usage in HBase

E
© Hortonworks Inc. 2011
HBase and HDFSUnderstanding file system usage in HBase
Enis Söztutar
enis [ at ] apache [dot] org
@enissoz
Page 1
© Hortonworks Inc. 2011
About Me
Page 2
Architecting the Future of Big Data
• In the Hadoop space since 2007
• Committer and PMC Member in Apache HBase and Hadoop
• Working at Hortonworks as member of Technical Staff
• Twitter: @enissoz
© Hortonworks Inc. 2011
Motivation
• HBase as a database depends on FileSystem for many things
• HBase has to work over HDFS, linux & windows
• HBase is the most advanced user of HDFS
• For tuning for IO performance, you have to understand how HBase does
IO
Page 3
Architecting the Future of Big Data
MapReduce
Large files
Few random seek
Batch oriented
High throughput
Failure handling at task level
Computation moves to data
HBase
Large files
A lot of random seek
Latency sensitive
Durability guarantees with sync
Computation generates local data
Large number of open files
© Hortonworks Inc. 2011
Agenda
• Overview of file types in Hbase
• Durability semantics
• IO Fencing / Lease recovery
• Data locality
– Short circuit reads (SSR)
– Checksums
– Block Placement
• Open topics
Page 4
Architecting the Future of Big Data
© Hortonworks Inc. 2011
HBase file types
Architecting the Future of Big Data
Page 5
© Hortonworks Inc. 2011
Overview of file types
• Mainly three types of files in Hbase
– Write Ahead Logs (a.k.a. WALs, logs)
– Data files (a.k.a. store files, hfiles)
– References / symbolic or logical links (0 length files)
• Every file is 3-way replicated
Page 6
Architecting the Future of Big Data
© Hortonworks Inc. 2011
Overview of file types
/hbase/.archive
/hbase/.logs/
/hbase/.logs/server1,60020,1370043265148/
/hbase/.logs/server1,60020,1370043265148/server1%2C60020%2C1370043265148.1370050467720
/hbase/.logs/server1,60020,1370043265105/server1%2C60020%2C1370043265105.1370046867591
…
/hbase/.oldlogs
/hbase/usertable/0711fd70ce0df641e9440e4979d67995/family/449e2fa173c14747b9d2e5..
/hbase/usertable/0711fd70ce0df641e9440e4979d67995/family/9103f38174ab48aa898a4b..
/hbase/table1/565bfb6220ca3edf02ac1f425cf18524/f1/49b32d3ee94543fb9055..
/hbase/.hbase-snapshot/usertable_snapshot/0ae3d2a93d3cf34a7cd30../family/12f114..
…
Page 7
Architecting the Future of Big Data
Write Ahead Logs
Data files
Links
© Hortonworks Inc. 2011
Data Files (HFile)
• Immutable once written
• Generated by flush or compactions (sequential writes)
• Read randomly (preads), or sequentially
• Big in size (flushsize -> tens of GBs)
• All data is in blocks (Hfile blocks not to be confused by HDFS blocks)
• Data blocks have target size:
– BLOCKSIZE in column family descriptor
– 64K by default
– Uncompressed and un-encoded size
• Index blocks (leaf, intermediate, root) have target size:
– hfile.index.block.max.size, 128K by default
• Bloom filter blocks have target size:
– io.storefile.bloom.block.size, 128K by default
Page 8
Architecting the Future of Big Data
© Hortonworks Inc. 2011
Data Files (HFile version 2.x)
Page 9
Architecting the Future of Big Data
© Hortonworks Inc. 2011
Data Files
• IO happens at block boundaries
– Random reads => disk seek + read whole block sequentially
– Read blocks are put into the block cache
– Leaf index blocks and bloom filter blocks also go to the block cache
• Use smaller block sizes for faster random-access
– Smaller read + faster in-block search
– Block index becomes bigger, more memory consumption
• Larger block sizes for faster scans
• Think about how many key values will fit in an average block
• Try compression and Data Block Encoding (PREFIX, DIFF, FAST_DIFF,
PREFIX_TREE)
– Minimizes file sizes + on disk block sizes
Page 10
Architecting the Future of Big Data
Key
length
Value
length
Row
length
Row key Family
length
Family Column
qualifier
Timesta
mp
KeyType Value
Int (4) Int (4) Short(2) Byte[] byte Byte[] Byte[] Long(8) byte Byte[]
© Hortonworks Inc. 2011
Reference Files / Links
• When region is split, “reference files” are created referring to the top or
bottom half of the parent store file according to splitkey
• HBase does not delete data/WAL files just “archives” them
/hbase/.oldlogs
/hbase/.archive
• Logs/hfiles are kept until TTL, and replication or snapshots are not
referring to them
– (hbase.master.logcleaner.ttl, 10min)
– (hbase.master.hfilecleaner.ttl, 5min)
• HFileLink: kind of hard / soft links that is application specific
• HBase snapshots are logical links to files (with backrefs)
Page 11
Architecting the Future of Big Data
© Hortonworks Inc. 2011
Write Ahead Logs
• One logical WAL per region / one physical per regionserver
• Rolled frequently
– hbase.regionserver.logroll.multiplier (0.95)
– hbase.regionserver.hlog.blocksize (default file system block size)
• Chronologically ordered set of files, only last one is open for writing
• Exceeding hbase.regionserver.maxlogs (32) will cause force flush
• Old log files are deleted as a whole
• Every edit is appended
• Sequential writes from WAL, sync very frequently (hundreds of times
per sec)
• Only sequential reads from replication, and crash recovery
• One log file per region server limits the write throughput per Region
Server
Page 12
Architecting the Future of Big Data
© Hortonworks Inc. 2011
Durability
(as in ACID)
Architecting the Future of Big Data
Page 13
© Hortonworks Inc. 2011
Overview of Write Path
1. Client sends the operations over RPC (Put/Delete)
2. Obtain row locks
3. Obtain the next mvcc write number
4. Tag the cells with the mvcc write number
5. Add the cells to the memstores (changes not visible yet)
6. Append a WALEdit to WAL, do not sync
7. Release row locks
8. Sync WAL
9. Advance mvcc, make changes visible
Page 14
Architecting the Future of Big Data
© Hortonworks Inc. 2011
Durability
• 0.94 and before:
– HTable property “DEFERRED_LOG_FLUSH” and
– Mutation.setWriteToWAL(false)
• 0.94 and 0.96:
Page 15
Architecting the Future of Big Data
Durability Semantics
USE_DEFAULT Use global hbase default, OR table default (SYNC_WAL)
SKIP_WAL Do not write updates to WAL
ASYNC_WAL Write entries to WAL asynchronously
(hbase.regionserver.optionallogflushinterval, 1 sec default)
SYNC_WAL Write entries to WAL, flush to datanodes
FSYNC_WAL Write entries to WAL, fsync in datanodes
© Hortonworks Inc. 2011
Durability
• 0.94 Durability setting per Mutation (HBASE-7801) / per table (HBASE-
8375)
• Allows intermixing different durability settings for updates to the same
table
• Durability is chosen from the mutation, unless it is USE_DEFAULT, in
which case Table’s Durability is used
• Limit the amount of time an edit can live in the memstore (HBASE-5930)
– hbase.regionserver.optionalcacheflushinterval
– Default 1hr
– Important for SKIP_WAL
– Cause a flush if there are unflushed edits that are older than
optionalcacheflushinterval
Page 16
Architecting the Future of Big Data
© Hortonworks Inc. 2011
Durability
Page 17
Architecting the Future of Big Data
public enum Durability {
USE_DEFAULT,
SKIP_WAL,
ASYNC_WAL,
SYNC_WAL,
FSYNC_WAL
}
Per Table:
HTableDescriptor htd = new HTableDescriptor("myTable");
htd.setDurability(Durability.ASYNC_WAL);
admin.createTable(htd);
Shell:
hbase(main):007:0> create 't12', 'f1', DURABILITY=>'ASYNC_WAL’
Per mutation:
Put put = new Put(rowKey);
put.setDurability(Durability.ASYNC_WAL);
table.put(put);
© Hortonworks Inc. 2011
Durability (Hflush / Hsync)
• Hflush() : Flush the data packet down the datanode pipeline. Wait for
ack’s.
• Hsync() : Flush the data packet down the pipeline. Have datanodes
execute FSYNC equivalent. Wait for ack’s.
• hflush is currently default, hsync() usage in HBase is not implemented
(HBASE-5954). Also not optimized (2x slow) and only Hadoop 2.0.
• hflush does not lose data, unless all 3 replicas die without syncing to
disk (datacenter power failure)
• Ensure that log is replicated 3 times
hbase.regionserver.hlog.tolerable.lowreplication
defaults to FileSystem default replication count (3 for HDFS)
Page 18
Architecting the Future of Big Data
public interface Syncable {
public void hflush() throws IOException;
public void hsync() throws IOException;
}
© Hortonworks Inc. 2011
Page 19
Architecting the Future of Big Data
© Hortonworks Inc. 2011
IO Fencing
Fencing is the process of isolating a node of a computer
cluster or protecting shared resources when a node appears
to be malfunctioning
Page 20
Architecting the Future of Big Data
© Hortonworks Inc. 2011
IO Fencing
Page 21
Architecting the Future of Big Data
Region1Client
Region Server A
(dying)
WAL
Region1
Region Server B
Append+sync
ack
edit
edit
WAL
Append+sync
ack
Master
Zookeeper
RegionServer A znode deleted
assign
Region1 Region Server A
Region 2 …
… …
YouAreDeadException
abort
RegionServer A session timeout
--
B
RegionServer A session timeout
Client
© Hortonworks Inc. 2011
IO Fencing
• Split Brain
• Ensure that a region is only hosted by a single region server at any time
• If master thinks that region server no longer hosts the region, RS
should not be able to accept and sync() updates
• Master renames the region server logs directory on HDFS:
– Current WAL cannot be rolled, new log file cannot be created
– For each WAL, before replaying recoverLease() is called
– recoverLease => lease recovery + block recovery
– Ensure that WAL is closed, and all data is visible (file length)
• Guarantees for region data files:
– Compactions => Remove files + add files
– Flushed => Allowed since resulting data is idempotent
• HBASE-2231, HBASE-7878, HBASE-8449
Page 22
Architecting the Future of Big Data
© Hortonworks Inc. 2011
Data Locality
Short circuit reads, checksums, block placement
Architecting the Future of Big Data
Page 23
© Hortonworks Inc. 2011
HDFS local reads (short circuit reads)
• Bypasses the datanode layer and directly
goes to the OS files
• Hadoop 1.x implementation:
– DFSClient asks for local paths for a block to the
local datanode
– Datanode checks whether the user has
permission
– Client gets the path for the block, opens the file
with FileInputStream
hdfs-site.xml
dfs.block.local-path-access.user = hbase
dfs.datanode.data.dir.perm = 750
hbase-site.xml
dfs.client.read.shortcircuit = true
Page 24
Architecting the Future of Big Data
RegionServer
Hadoop FileSystem
DFSClient
Datanode
OS Filesystem (ext3)
Disks
Disks
Disks
HBase Client
RPC
RPC
BlockReader
© Hortonworks Inc. 2011
HDFS local reads (short circuit reads)
• Hadoop 2.0 implementation (HDFS-347)
– Keep the legacy implementation
– Use Unix Domain sockets to pass the File Descriptor (FD)
– Datanode opens the block file and passes FD to the BlockReaderLocal running in
Regionserver process
– More secure than previous implementation
– Windows also supports domain sockets, need to implement native APIs
• Local buffer size dfs.client.read.shortcircuit.buffer.size
– BlockReaderLocal will fill this whole buffer everytime HBase will try to read an
HfileBlock
– dfs.client.read.shortcircuit.buffer.size = 1MB vs 64KB Hfile block size
– SSR buffer is a direct buffer (in Hadoop 2, not in Hadoop 1)
– # regions x # stores x #avg store files x # avg blocks per file x SSR buffer size
– 10 regions x 2 x 4 x (1GB / 64MB) x 1 MB = 1.28GB
non-heap memory usage
Page 25
Architecting the Future of Big Data
© Hortonworks Inc. 2011
Checksums
• HDFS checksums are not inlined.
• Two files per block, one for data, one for
checksums (HDFS-2699)
• Random positioned read causes 2 seeks
• HBase checksums comes with 0.94 (HDP
1.2+). HBASE-5074.
Page 26
Architecting the Future of Big Data
blk_123456789
.blk_123456789.meta
: Data chunk (dfs.bytes-per-checksum, 512 bytes)
: Checksum chunk (4 bytes)
© Hortonworks Inc. 2011
Checksums
Page 27
Architecting the Future of Big Data
• HFile version 2.1 writes checksums per
Hfile block
• HDFS checksum verification is bypassed
on block read, will be done by HBase
• If checksum fails, we go back to reading
checksums from HDFS for “some time”
• Due to double checksum bug(HDFS-3429)
in remote reads in Hadoop 1, not enabled
by default for now. Benchmark it yourself
hbase.regionserver.checksum.verify = true
hbase.hstore.bytes.per.checksum = 16384
hbase.hstore.checksum.algorithm = CRC32C
Never set this:
dfs.client.read.shortcircuit.skip.checksum = false
HFile
: Hfile data block chunk
: Checksum chunk
Hfile block
: Block header
© Hortonworks Inc. 2011
Rack 1 / Server 1
DataNode
Default Block Placement Policy
Page 28
Architecting the Future of Big Data
b1
RegionServer
Region A
Region B
StoreFile
StoreFile
StoreFile
StoreFile
StoreFile
b2 b2
b9 b1
b1
b2
b3
b2
b1 b2b1
Rack N / Server M
DataNode
b2
b1
b1
Rack L / Server K
DataNode
b2
b1
Rack X / Server Y
DataNode
b1b2 b2
b3
RegionServer RegionServer RegionServer
© Hortonworks Inc. 2011
Data locality for HBase
• Poor data locality when the region is moved:
– As a result of load balancing
– Region server crash + failover
• Most of the data won’t be local unless the files are compacted
• Idea (from Facebook): Regions have affiliated nodes (primary,
secondary, tertiary), HBASE-4755
• When writing a data file, give hints to the NN that we want these
locations for block replicas (HDFS-2576)
• LB should assign the region to one of the affiliated nodes on server
crash
– Keep data locality
– SSR will still work
• Reduces data loss probability
Page 29
Architecting the Future of Big Data
© Hortonworks Inc. 2011
Rack X / Server Y
RegionServer
Rack L / Server K
RegionServer
Rack N / Server M
RegionServer
Rack 1 / Server 1
Default Block Placement Policy
Page 30
Architecting the Future of Big Data
RegionServer
Region A
StoreFile
StoreFile
StoreFile
Region B
StoreFile
StoreFile
DataNode
b1
b2 b2
b9 b1
b1
b2
b3
b2
b1 b2b1
DataNode
b1
b2
b2
b9b1
b2
b1
DataNode
b1
b2
b2
b9
b2
b1
DataNode
b1
b2
b3
b2
b1
© Hortonworks Inc. 2011
Other considerations
• HBase riding over Namenode HA
– Both Hadoop 1 (NFS based) and Hadoop 2 HA (JQM, etc)
– Heavily tested with full stack HA
• Retry HDFS operations
• Isolate FileSystem usage from HBase internals
• Hadoop 2 vs Hadoop 1 performance
– Hadoop 2 is coming!
• HDFS snapshots vs HBase snapshots
– HBase DOES NOT use HDFS snapshots
– Need hardlinks
– Super flush API
• HBase security vs HDFS security
– All files are owned by HBase principal
– No ACL’s in HDFS. Allowing a user to read Hfiles / snapshots directly is hard
Page 31
Architecting the Future of Big Data
© Hortonworks Inc. 2011
Open Topics
• HDFS hard links
– Rethink how we do snapshots, backups, etc
• Parallel writes for WAL
– Reduce latency on WAL syncs
• SSD storage, cache
– SSD storage type in Hadoop or local filesystem
– Using SSD’s as a secondary cache
– Selectively places tables / column families on SSD
• HDFS zero-copy reads (HDFS-3051, HADOOP-8148)
• HDFS inline checksums (HDFS-2699)
• HDFS Quorum reads (HBASE-7509)
Page 32
Architecting the Future of Big Data
© Hortonworks Inc. 2011
Thanks
Questions?
Architecting the Future of Big Data
Page 33
Enis Söztutar
enis [ at ] apache [dot] org
@enissoz
1 of 33

Recommended

What is new in Apache Hive 3.0? by
What is new in Apache Hive 3.0?What is new in Apache Hive 3.0?
What is new in Apache Hive 3.0?DataWorks Summit
6.6K views32 slides
Performance Optimizations in Apache Impala by
Performance Optimizations in Apache ImpalaPerformance Optimizations in Apache Impala
Performance Optimizations in Apache ImpalaCloudera, Inc.
10.7K views63 slides
Introduction to memcached by
Introduction to memcachedIntroduction to memcached
Introduction to memcachedJurriaan Persyn
70.9K views77 slides
HBaseCon 2015: HBase Performance Tuning @ Salesforce by
HBaseCon 2015: HBase Performance Tuning @ SalesforceHBaseCon 2015: HBase Performance Tuning @ Salesforce
HBaseCon 2015: HBase Performance Tuning @ SalesforceHBaseCon
6.1K views54 slides
RocksDB detail by
RocksDB detailRocksDB detail
RocksDB detailMIJIN AN
7.3K views27 slides
Apache Spark Architecture by
Apache Spark ArchitectureApache Spark Architecture
Apache Spark ArchitectureAlexey Grishchenko
76K views114 slides

More Related Content

What's hot

Power of the Log: LSM & Append Only Data Structures by
Power of the Log: LSM & Append Only Data StructuresPower of the Log: LSM & Append Only Data Structures
Power of the Log: LSM & Append Only Data Structuresconfluent
6.8K views71 slides
Introduction and Overview of Apache Kafka, TriHUG July 23, 2013 by
Introduction and Overview of Apache Kafka, TriHUG July 23, 2013Introduction and Overview of Apache Kafka, TriHUG July 23, 2013
Introduction and Overview of Apache Kafka, TriHUG July 23, 2013mumrah
61.3K views43 slides
Kafka replication apachecon_2013 by
Kafka replication apachecon_2013Kafka replication apachecon_2013
Kafka replication apachecon_2013Jun Rao
21.3K views31 slides
Hudi: Large-Scale, Near Real-Time Pipelines at Uber with Nishith Agarwal and ... by
Hudi: Large-Scale, Near Real-Time Pipelines at Uber with Nishith Agarwal and ...Hudi: Large-Scale, Near Real-Time Pipelines at Uber with Nishith Agarwal and ...
Hudi: Large-Scale, Near Real-Time Pipelines at Uber with Nishith Agarwal and ...Databricks
3.8K views28 slides
HBase Low Latency by
HBase Low LatencyHBase Low Latency
HBase Low LatencyDataWorks Summit
5.1K views38 slides
Hive + Tez: A Performance Deep Dive by
Hive + Tez: A Performance Deep DiveHive + Tez: A Performance Deep Dive
Hive + Tez: A Performance Deep DiveDataWorks Summit
57.6K views41 slides

What's hot(20)

Power of the Log: LSM & Append Only Data Structures by confluent
Power of the Log: LSM & Append Only Data StructuresPower of the Log: LSM & Append Only Data Structures
Power of the Log: LSM & Append Only Data Structures
confluent6.8K views
Introduction and Overview of Apache Kafka, TriHUG July 23, 2013 by mumrah
Introduction and Overview of Apache Kafka, TriHUG July 23, 2013Introduction and Overview of Apache Kafka, TriHUG July 23, 2013
Introduction and Overview of Apache Kafka, TriHUG July 23, 2013
mumrah61.3K views
Kafka replication apachecon_2013 by Jun Rao
Kafka replication apachecon_2013Kafka replication apachecon_2013
Kafka replication apachecon_2013
Jun Rao21.3K views
Hudi: Large-Scale, Near Real-Time Pipelines at Uber with Nishith Agarwal and ... by Databricks
Hudi: Large-Scale, Near Real-Time Pipelines at Uber with Nishith Agarwal and ...Hudi: Large-Scale, Near Real-Time Pipelines at Uber with Nishith Agarwal and ...
Hudi: Large-Scale, Near Real-Time Pipelines at Uber with Nishith Agarwal and ...
Databricks3.8K views
Hive + Tez: A Performance Deep Dive by DataWorks Summit
Hive + Tez: A Performance Deep DiveHive + Tez: A Performance Deep Dive
Hive + Tez: A Performance Deep Dive
DataWorks Summit57.6K views
Deep Dive: Memory Management in Apache Spark by Databricks
Deep Dive: Memory Management in Apache SparkDeep Dive: Memory Management in Apache Spark
Deep Dive: Memory Management in Apache Spark
Databricks14.5K views
Amazon S3 Best Practice and Tuning for Hadoop/Spark in the Cloud by Noritaka Sekiyama
Amazon S3 Best Practice and Tuning for Hadoop/Spark in the CloudAmazon S3 Best Practice and Tuning for Hadoop/Spark in the Cloud
Amazon S3 Best Practice and Tuning for Hadoop/Spark in the Cloud
Noritaka Sekiyama33.3K views
Improving SparkSQL Performance by 30%: How We Optimize Parquet Pushdown and P... by Databricks
Improving SparkSQL Performance by 30%: How We Optimize Parquet Pushdown and P...Improving SparkSQL Performance by 30%: How We Optimize Parquet Pushdown and P...
Improving SparkSQL Performance by 30%: How We Optimize Parquet Pushdown and P...
Databricks1K views
Redis + Apache Spark = Swiss Army Knife Meets Kitchen Sink by Databricks
Redis + Apache Spark = Swiss Army Knife Meets Kitchen SinkRedis + Apache Spark = Swiss Army Knife Meets Kitchen Sink
Redis + Apache Spark = Swiss Army Knife Meets Kitchen Sink
Databricks677 views
Iceberg + Alluxio for Fast Data Analytics by Alluxio, Inc.
Iceberg + Alluxio for Fast Data AnalyticsIceberg + Alluxio for Fast Data Analytics
Iceberg + Alluxio for Fast Data Analytics
Alluxio, Inc.530 views
Thousands of Threads and Blocking I/O by George Cao
Thousands of Threads and Blocking I/OThousands of Threads and Blocking I/O
Thousands of Threads and Blocking I/O
George Cao17.1K views
Apache HBase Performance Tuning by Lars Hofhansl
Apache HBase Performance TuningApache HBase Performance Tuning
Apache HBase Performance Tuning
Lars Hofhansl39.6K views
Introduction to Apache Flink - Fast and reliable big data processing by Till Rohrmann
Introduction to Apache Flink - Fast and reliable big data processingIntroduction to Apache Flink - Fast and reliable big data processing
Introduction to Apache Flink - Fast and reliable big data processing
Till Rohrmann7.2K views
Facebook Messages & HBase by 强 王
Facebook Messages & HBaseFacebook Messages & HBase
Facebook Messages & HBase
强 王39.2K views
Introduction to Apache ZooKeeper by Saurav Haloi
Introduction to Apache ZooKeeperIntroduction to Apache ZooKeeper
Introduction to Apache ZooKeeper
Saurav Haloi128.6K views
A Thorough Comparison of Delta Lake, Iceberg and Hudi by Databricks
A Thorough Comparison of Delta Lake, Iceberg and HudiA Thorough Comparison of Delta Lake, Iceberg and Hudi
A Thorough Comparison of Delta Lake, Iceberg and Hudi
Databricks11.1K views
Introduction to Redis by Dvir Volk
Introduction to RedisIntroduction to Redis
Introduction to Redis
Dvir Volk121.1K views

Similar to HBase and HDFS: Understanding FileSystem Usage in HBase

Ozone and HDFS's Evolution by
Ozone and HDFS's EvolutionOzone and HDFS's Evolution
Ozone and HDFS's EvolutionDataWorks Summit
991 views28 slides
Ozone and HDFS’s evolution by
Ozone and HDFS’s evolutionOzone and HDFS’s evolution
Ozone and HDFS’s evolutionDataWorks Summit
198 views28 slides
Evolving HDFS to a Generalized Storage Subsystem by
Evolving HDFS to a Generalized Storage SubsystemEvolving HDFS to a Generalized Storage Subsystem
Evolving HDFS to a Generalized Storage SubsystemDataWorks Summit/Hadoop Summit
1.5K views22 slides
Ozone and HDFS’s evolution by
Ozone and HDFS’s evolutionOzone and HDFS’s evolution
Ozone and HDFS’s evolutionDataWorks Summit
2.4K views29 slides
HBaseCon 2013: Compaction Improvements in Apache HBase by
HBaseCon 2013: Compaction Improvements in Apache HBaseHBaseCon 2013: Compaction Improvements in Apache HBase
HBaseCon 2013: Compaction Improvements in Apache HBaseCloudera, Inc.
19.1K views33 slides
HBase for Architects by
HBase for ArchitectsHBase for Architects
HBase for ArchitectsNick Dimiduk
33.7K views21 slides

Similar to HBase and HDFS: Understanding FileSystem Usage in HBase(20)

HBaseCon 2013: Compaction Improvements in Apache HBase by Cloudera, Inc.
HBaseCon 2013: Compaction Improvements in Apache HBaseHBaseCon 2013: Compaction Improvements in Apache HBase
HBaseCon 2013: Compaction Improvements in Apache HBase
Cloudera, Inc.19.1K views
HBase for Architects by Nick Dimiduk
HBase for ArchitectsHBase for Architects
HBase for Architects
Nick Dimiduk33.7K views
Meet hbase 2.0 by enissoz
Meet hbase 2.0Meet hbase 2.0
Meet hbase 2.0
enissoz5.3K views
Meet HBase 2.0 by enissoz
Meet HBase 2.0Meet HBase 2.0
Meet HBase 2.0
enissoz675 views
Big Data and Hadoop - History, Technical Deep Dive, and Industry Trends by Esther Kundin
Big Data and Hadoop - History, Technical Deep Dive, and Industry TrendsBig Data and Hadoop - History, Technical Deep Dive, and Industry Trends
Big Data and Hadoop - History, Technical Deep Dive, and Industry Trends
Esther Kundin1.1K views
[B4]deview 2012-hdfs by NAVER D2
[B4]deview 2012-hdfs[B4]deview 2012-hdfs
[B4]deview 2012-hdfs
NAVER D21.9K views
Apache HBase Internals you hoped you Never Needed to Understand by Josh Elser
Apache HBase Internals you hoped you Never Needed to UnderstandApache HBase Internals you hoped you Never Needed to Understand
Apache HBase Internals you hoped you Never Needed to Understand
Josh Elser2.1K views
Storage and-compute-hdfs-map reduce by Chris Nauroth
Storage and-compute-hdfs-map reduceStorage and-compute-hdfs-map reduce
Storage and-compute-hdfs-map reduce
Chris Nauroth655 views
Big Data and Hadoop - History, Technical Deep Dive, and Industry Trends by Esther Kundin
Big Data and Hadoop - History, Technical Deep Dive, and Industry TrendsBig Data and Hadoop - History, Technical Deep Dive, and Industry Trends
Big Data and Hadoop - History, Technical Deep Dive, and Industry Trends
Esther Kundin564 views
Big data processing engines, Atlanta Meetup 4/30 by Ashish Narasimham
Big data processing engines, Atlanta Meetup 4/30Big data processing engines, Atlanta Meetup 4/30
Big data processing engines, Atlanta Meetup 4/30
Large-scale Web Apps @ Pinterest by HBaseCon
Large-scale Web Apps @ PinterestLarge-scale Web Apps @ Pinterest
Large-scale Web Apps @ Pinterest
HBaseCon4.1K views

More from enissoz

Operating and supporting HBase Clusters by
Operating and supporting HBase ClustersOperating and supporting HBase Clusters
Operating and supporting HBase Clustersenissoz
1.4K views43 slides
HBase state of the union by
HBase   state of the unionHBase   state of the union
HBase state of the unionenissoz
552 views29 slides
Apache phoenix: Past, Present and Future of SQL over HBAse by
Apache phoenix: Past, Present and Future of SQL over HBAseApache phoenix: Past, Present and Future of SQL over HBAse
Apache phoenix: Past, Present and Future of SQL over HBAseenissoz
6.3K views41 slides
Meet HBase 1.0 by
Meet HBase 1.0Meet HBase 1.0
Meet HBase 1.0enissoz
8.2K views48 slides
HBase Read High Availability Using Timeline Consistent Region Replicas by
HBase  Read High Availability Using Timeline Consistent Region ReplicasHBase  Read High Availability Using Timeline Consistent Region Replicas
HBase Read High Availability Using Timeline Consistent Region Replicasenissoz
8.7K views38 slides
Mapreduce over snapshots by
Mapreduce over snapshotsMapreduce over snapshots
Mapreduce over snapshotsenissoz
6.4K views16 slides

More from enissoz(6)

Operating and supporting HBase Clusters by enissoz
Operating and supporting HBase ClustersOperating and supporting HBase Clusters
Operating and supporting HBase Clusters
enissoz1.4K views
HBase state of the union by enissoz
HBase   state of the unionHBase   state of the union
HBase state of the union
enissoz552 views
Apache phoenix: Past, Present and Future of SQL over HBAse by enissoz
Apache phoenix: Past, Present and Future of SQL over HBAseApache phoenix: Past, Present and Future of SQL over HBAse
Apache phoenix: Past, Present and Future of SQL over HBAse
enissoz6.3K views
Meet HBase 1.0 by enissoz
Meet HBase 1.0Meet HBase 1.0
Meet HBase 1.0
enissoz8.2K views
HBase Read High Availability Using Timeline Consistent Region Replicas by enissoz
HBase  Read High Availability Using Timeline Consistent Region ReplicasHBase  Read High Availability Using Timeline Consistent Region Replicas
HBase Read High Availability Using Timeline Consistent Region Replicas
enissoz8.7K views
Mapreduce over snapshots by enissoz
Mapreduce over snapshotsMapreduce over snapshots
Mapreduce over snapshots
enissoz6.4K views

Recently uploaded

Inawisdom IDP by
Inawisdom IDPInawisdom IDP
Inawisdom IDPPhilipBasford
17 views48 slides
KubeConNA23 Recap.pdf by
KubeConNA23 Recap.pdfKubeConNA23 Recap.pdf
KubeConNA23 Recap.pdfMichaelOLeary82
28 views27 slides
MVP and prioritization.pdf by
MVP and prioritization.pdfMVP and prioritization.pdf
MVP and prioritization.pdfrahuldharwal141
40 views8 slides
PCCC23:日本AMD株式会社 テーマ1「AMD Instinct™ アクセラレーターの概要」 by
PCCC23:日本AMD株式会社 テーマ1「AMD Instinct™ アクセラレーターの概要」PCCC23:日本AMD株式会社 テーマ1「AMD Instinct™ アクセラレーターの概要」
PCCC23:日本AMD株式会社 テーマ1「AMD Instinct™ アクセラレーターの概要」PC Cluster Consortium
29 views68 slides
Initiating and Advancing Your Strategic GIS Governance Strategy by
Initiating and Advancing Your Strategic GIS Governance StrategyInitiating and Advancing Your Strategic GIS Governance Strategy
Initiating and Advancing Your Strategic GIS Governance StrategySafe Software
198 views68 slides
Cocktail of Environments. How to Mix Test and Development Environments and St... by
Cocktail of Environments. How to Mix Test and Development Environments and St...Cocktail of Environments. How to Mix Test and Development Environments and St...
Cocktail of Environments. How to Mix Test and Development Environments and St...Aleksandr Tarasov
26 views135 slides

Recently uploaded(20)

PCCC23:日本AMD株式会社 テーマ1「AMD Instinct™ アクセラレーターの概要」 by PC Cluster Consortium
PCCC23:日本AMD株式会社 テーマ1「AMD Instinct™ アクセラレーターの概要」PCCC23:日本AMD株式会社 テーマ1「AMD Instinct™ アクセラレーターの概要」
PCCC23:日本AMD株式会社 テーマ1「AMD Instinct™ アクセラレーターの概要」
Initiating and Advancing Your Strategic GIS Governance Strategy by Safe Software
Initiating and Advancing Your Strategic GIS Governance StrategyInitiating and Advancing Your Strategic GIS Governance Strategy
Initiating and Advancing Your Strategic GIS Governance Strategy
Safe Software198 views
Cocktail of Environments. How to Mix Test and Development Environments and St... by Aleksandr Tarasov
Cocktail of Environments. How to Mix Test and Development Environments and St...Cocktail of Environments. How to Mix Test and Development Environments and St...
Cocktail of Environments. How to Mix Test and Development Environments and St...
Enabling DPU Hardware Accelerators in XCP-ng Cloud Platform Environment - And... by ShapeBlue
Enabling DPU Hardware Accelerators in XCP-ng Cloud Platform Environment - And...Enabling DPU Hardware Accelerators in XCP-ng Cloud Platform Environment - And...
Enabling DPU Hardware Accelerators in XCP-ng Cloud Platform Environment - And...
ShapeBlue120 views
PCCC23:日本AMD株式会社 テーマ2「AMD EPYC™ プロセッサーを用いたAIソリューション」 by PC Cluster Consortium
PCCC23:日本AMD株式会社 テーマ2「AMD EPYC™ プロセッサーを用いたAIソリューション」PCCC23:日本AMD株式会社 テーマ2「AMD EPYC™ プロセッサーを用いたAIソリューション」
PCCC23:日本AMD株式会社 テーマ2「AMD EPYC™ プロセッサーを用いたAIソリューション」
Digital Personal Data Protection (DPDP) Practical Approach For CISOs by Priyanka Aash
Digital Personal Data Protection (DPDP) Practical Approach For CISOsDigital Personal Data Protection (DPDP) Practical Approach For CISOs
Digital Personal Data Protection (DPDP) Practical Approach For CISOs
Priyanka Aash171 views
"Surviving highload with Node.js", Andrii Shumada by Fwdays
"Surviving highload with Node.js", Andrii Shumada "Surviving highload with Node.js", Andrii Shumada
"Surviving highload with Node.js", Andrii Shumada
Fwdays59 views
Future of AR - Facebook Presentation by Rob McCarty
Future of AR - Facebook PresentationFuture of AR - Facebook Presentation
Future of AR - Facebook Presentation
Rob McCarty66 views
Deep Tech and the Amplified Organisation: Core Concepts by Holonomics
Deep Tech and the Amplified Organisation: Core ConceptsDeep Tech and the Amplified Organisation: Core Concepts
Deep Tech and the Amplified Organisation: Core Concepts
Holonomics17 views
"Node.js vs workers — A comparison of two JavaScript runtimes", James M Snell by Fwdays
"Node.js vs workers — A comparison of two JavaScript runtimes", James M Snell"Node.js vs workers — A comparison of two JavaScript runtimes", James M Snell
"Node.js vs workers — A comparison of two JavaScript runtimes", James M Snell
Fwdays14 views
"Running students' code in isolation. The hard way", Yurii Holiuk by Fwdays
"Running students' code in isolation. The hard way", Yurii Holiuk "Running students' code in isolation. The hard way", Yurii Holiuk
"Running students' code in isolation. The hard way", Yurii Holiuk
Fwdays38 views
Measurecamp Brussels - Synthetic data.pdf by Human37
Measurecamp Brussels - Synthetic data.pdfMeasurecamp Brussels - Synthetic data.pdf
Measurecamp Brussels - Synthetic data.pdf
Human37 27 views
Discover Aura Workshop (12.5.23).pdf by Neo4j
Discover Aura Workshop (12.5.23).pdfDiscover Aura Workshop (12.5.23).pdf
Discover Aura Workshop (12.5.23).pdf
Neo4j20 views
AIM102-S_Cognizant_CognizantCognitive by PhilipBasford
AIM102-S_Cognizant_CognizantCognitiveAIM102-S_Cognizant_CognizantCognitive
AIM102-S_Cognizant_CognizantCognitive
PhilipBasford23 views
The Coming AI Tsunami.pptx by johnhandby
The Coming AI Tsunami.pptxThe Coming AI Tsunami.pptx
The Coming AI Tsunami.pptx
johnhandby14 views

HBase and HDFS: Understanding FileSystem Usage in HBase

  • 1. © Hortonworks Inc. 2011 HBase and HDFSUnderstanding file system usage in HBase Enis Söztutar enis [ at ] apache [dot] org @enissoz Page 1
  • 2. © Hortonworks Inc. 2011 About Me Page 2 Architecting the Future of Big Data • In the Hadoop space since 2007 • Committer and PMC Member in Apache HBase and Hadoop • Working at Hortonworks as member of Technical Staff • Twitter: @enissoz
  • 3. © Hortonworks Inc. 2011 Motivation • HBase as a database depends on FileSystem for many things • HBase has to work over HDFS, linux & windows • HBase is the most advanced user of HDFS • For tuning for IO performance, you have to understand how HBase does IO Page 3 Architecting the Future of Big Data MapReduce Large files Few random seek Batch oriented High throughput Failure handling at task level Computation moves to data HBase Large files A lot of random seek Latency sensitive Durability guarantees with sync Computation generates local data Large number of open files
  • 4. © Hortonworks Inc. 2011 Agenda • Overview of file types in Hbase • Durability semantics • IO Fencing / Lease recovery • Data locality – Short circuit reads (SSR) – Checksums – Block Placement • Open topics Page 4 Architecting the Future of Big Data
  • 5. © Hortonworks Inc. 2011 HBase file types Architecting the Future of Big Data Page 5
  • 6. © Hortonworks Inc. 2011 Overview of file types • Mainly three types of files in Hbase – Write Ahead Logs (a.k.a. WALs, logs) – Data files (a.k.a. store files, hfiles) – References / symbolic or logical links (0 length files) • Every file is 3-way replicated Page 6 Architecting the Future of Big Data
  • 7. © Hortonworks Inc. 2011 Overview of file types /hbase/.archive /hbase/.logs/ /hbase/.logs/server1,60020,1370043265148/ /hbase/.logs/server1,60020,1370043265148/server1%2C60020%2C1370043265148.1370050467720 /hbase/.logs/server1,60020,1370043265105/server1%2C60020%2C1370043265105.1370046867591 … /hbase/.oldlogs /hbase/usertable/0711fd70ce0df641e9440e4979d67995/family/449e2fa173c14747b9d2e5.. /hbase/usertable/0711fd70ce0df641e9440e4979d67995/family/9103f38174ab48aa898a4b.. /hbase/table1/565bfb6220ca3edf02ac1f425cf18524/f1/49b32d3ee94543fb9055.. /hbase/.hbase-snapshot/usertable_snapshot/0ae3d2a93d3cf34a7cd30../family/12f114.. … Page 7 Architecting the Future of Big Data Write Ahead Logs Data files Links
  • 8. © Hortonworks Inc. 2011 Data Files (HFile) • Immutable once written • Generated by flush or compactions (sequential writes) • Read randomly (preads), or sequentially • Big in size (flushsize -> tens of GBs) • All data is in blocks (Hfile blocks not to be confused by HDFS blocks) • Data blocks have target size: – BLOCKSIZE in column family descriptor – 64K by default – Uncompressed and un-encoded size • Index blocks (leaf, intermediate, root) have target size: – hfile.index.block.max.size, 128K by default • Bloom filter blocks have target size: – io.storefile.bloom.block.size, 128K by default Page 8 Architecting the Future of Big Data
  • 9. © Hortonworks Inc. 2011 Data Files (HFile version 2.x) Page 9 Architecting the Future of Big Data
  • 10. © Hortonworks Inc. 2011 Data Files • IO happens at block boundaries – Random reads => disk seek + read whole block sequentially – Read blocks are put into the block cache – Leaf index blocks and bloom filter blocks also go to the block cache • Use smaller block sizes for faster random-access – Smaller read + faster in-block search – Block index becomes bigger, more memory consumption • Larger block sizes for faster scans • Think about how many key values will fit in an average block • Try compression and Data Block Encoding (PREFIX, DIFF, FAST_DIFF, PREFIX_TREE) – Minimizes file sizes + on disk block sizes Page 10 Architecting the Future of Big Data Key length Value length Row length Row key Family length Family Column qualifier Timesta mp KeyType Value Int (4) Int (4) Short(2) Byte[] byte Byte[] Byte[] Long(8) byte Byte[]
  • 11. © Hortonworks Inc. 2011 Reference Files / Links • When region is split, “reference files” are created referring to the top or bottom half of the parent store file according to splitkey • HBase does not delete data/WAL files just “archives” them /hbase/.oldlogs /hbase/.archive • Logs/hfiles are kept until TTL, and replication or snapshots are not referring to them – (hbase.master.logcleaner.ttl, 10min) – (hbase.master.hfilecleaner.ttl, 5min) • HFileLink: kind of hard / soft links that is application specific • HBase snapshots are logical links to files (with backrefs) Page 11 Architecting the Future of Big Data
  • 12. © Hortonworks Inc. 2011 Write Ahead Logs • One logical WAL per region / one physical per regionserver • Rolled frequently – hbase.regionserver.logroll.multiplier (0.95) – hbase.regionserver.hlog.blocksize (default file system block size) • Chronologically ordered set of files, only last one is open for writing • Exceeding hbase.regionserver.maxlogs (32) will cause force flush • Old log files are deleted as a whole • Every edit is appended • Sequential writes from WAL, sync very frequently (hundreds of times per sec) • Only sequential reads from replication, and crash recovery • One log file per region server limits the write throughput per Region Server Page 12 Architecting the Future of Big Data
  • 13. © Hortonworks Inc. 2011 Durability (as in ACID) Architecting the Future of Big Data Page 13
  • 14. © Hortonworks Inc. 2011 Overview of Write Path 1. Client sends the operations over RPC (Put/Delete) 2. Obtain row locks 3. Obtain the next mvcc write number 4. Tag the cells with the mvcc write number 5. Add the cells to the memstores (changes not visible yet) 6. Append a WALEdit to WAL, do not sync 7. Release row locks 8. Sync WAL 9. Advance mvcc, make changes visible Page 14 Architecting the Future of Big Data
  • 15. © Hortonworks Inc. 2011 Durability • 0.94 and before: – HTable property “DEFERRED_LOG_FLUSH” and – Mutation.setWriteToWAL(false) • 0.94 and 0.96: Page 15 Architecting the Future of Big Data Durability Semantics USE_DEFAULT Use global hbase default, OR table default (SYNC_WAL) SKIP_WAL Do not write updates to WAL ASYNC_WAL Write entries to WAL asynchronously (hbase.regionserver.optionallogflushinterval, 1 sec default) SYNC_WAL Write entries to WAL, flush to datanodes FSYNC_WAL Write entries to WAL, fsync in datanodes
  • 16. © Hortonworks Inc. 2011 Durability • 0.94 Durability setting per Mutation (HBASE-7801) / per table (HBASE- 8375) • Allows intermixing different durability settings for updates to the same table • Durability is chosen from the mutation, unless it is USE_DEFAULT, in which case Table’s Durability is used • Limit the amount of time an edit can live in the memstore (HBASE-5930) – hbase.regionserver.optionalcacheflushinterval – Default 1hr – Important for SKIP_WAL – Cause a flush if there are unflushed edits that are older than optionalcacheflushinterval Page 16 Architecting the Future of Big Data
  • 17. © Hortonworks Inc. 2011 Durability Page 17 Architecting the Future of Big Data public enum Durability { USE_DEFAULT, SKIP_WAL, ASYNC_WAL, SYNC_WAL, FSYNC_WAL } Per Table: HTableDescriptor htd = new HTableDescriptor("myTable"); htd.setDurability(Durability.ASYNC_WAL); admin.createTable(htd); Shell: hbase(main):007:0> create 't12', 'f1', DURABILITY=>'ASYNC_WAL’ Per mutation: Put put = new Put(rowKey); put.setDurability(Durability.ASYNC_WAL); table.put(put);
  • 18. © Hortonworks Inc. 2011 Durability (Hflush / Hsync) • Hflush() : Flush the data packet down the datanode pipeline. Wait for ack’s. • Hsync() : Flush the data packet down the pipeline. Have datanodes execute FSYNC equivalent. Wait for ack’s. • hflush is currently default, hsync() usage in HBase is not implemented (HBASE-5954). Also not optimized (2x slow) and only Hadoop 2.0. • hflush does not lose data, unless all 3 replicas die without syncing to disk (datacenter power failure) • Ensure that log is replicated 3 times hbase.regionserver.hlog.tolerable.lowreplication defaults to FileSystem default replication count (3 for HDFS) Page 18 Architecting the Future of Big Data public interface Syncable { public void hflush() throws IOException; public void hsync() throws IOException; }
  • 19. © Hortonworks Inc. 2011 Page 19 Architecting the Future of Big Data
  • 20. © Hortonworks Inc. 2011 IO Fencing Fencing is the process of isolating a node of a computer cluster or protecting shared resources when a node appears to be malfunctioning Page 20 Architecting the Future of Big Data
  • 21. © Hortonworks Inc. 2011 IO Fencing Page 21 Architecting the Future of Big Data Region1Client Region Server A (dying) WAL Region1 Region Server B Append+sync ack edit edit WAL Append+sync ack Master Zookeeper RegionServer A znode deleted assign Region1 Region Server A Region 2 … … … YouAreDeadException abort RegionServer A session timeout -- B RegionServer A session timeout Client
  • 22. © Hortonworks Inc. 2011 IO Fencing • Split Brain • Ensure that a region is only hosted by a single region server at any time • If master thinks that region server no longer hosts the region, RS should not be able to accept and sync() updates • Master renames the region server logs directory on HDFS: – Current WAL cannot be rolled, new log file cannot be created – For each WAL, before replaying recoverLease() is called – recoverLease => lease recovery + block recovery – Ensure that WAL is closed, and all data is visible (file length) • Guarantees for region data files: – Compactions => Remove files + add files – Flushed => Allowed since resulting data is idempotent • HBASE-2231, HBASE-7878, HBASE-8449 Page 22 Architecting the Future of Big Data
  • 23. © Hortonworks Inc. 2011 Data Locality Short circuit reads, checksums, block placement Architecting the Future of Big Data Page 23
  • 24. © Hortonworks Inc. 2011 HDFS local reads (short circuit reads) • Bypasses the datanode layer and directly goes to the OS files • Hadoop 1.x implementation: – DFSClient asks for local paths for a block to the local datanode – Datanode checks whether the user has permission – Client gets the path for the block, opens the file with FileInputStream hdfs-site.xml dfs.block.local-path-access.user = hbase dfs.datanode.data.dir.perm = 750 hbase-site.xml dfs.client.read.shortcircuit = true Page 24 Architecting the Future of Big Data RegionServer Hadoop FileSystem DFSClient Datanode OS Filesystem (ext3) Disks Disks Disks HBase Client RPC RPC BlockReader
  • 25. © Hortonworks Inc. 2011 HDFS local reads (short circuit reads) • Hadoop 2.0 implementation (HDFS-347) – Keep the legacy implementation – Use Unix Domain sockets to pass the File Descriptor (FD) – Datanode opens the block file and passes FD to the BlockReaderLocal running in Regionserver process – More secure than previous implementation – Windows also supports domain sockets, need to implement native APIs • Local buffer size dfs.client.read.shortcircuit.buffer.size – BlockReaderLocal will fill this whole buffer everytime HBase will try to read an HfileBlock – dfs.client.read.shortcircuit.buffer.size = 1MB vs 64KB Hfile block size – SSR buffer is a direct buffer (in Hadoop 2, not in Hadoop 1) – # regions x # stores x #avg store files x # avg blocks per file x SSR buffer size – 10 regions x 2 x 4 x (1GB / 64MB) x 1 MB = 1.28GB non-heap memory usage Page 25 Architecting the Future of Big Data
  • 26. © Hortonworks Inc. 2011 Checksums • HDFS checksums are not inlined. • Two files per block, one for data, one for checksums (HDFS-2699) • Random positioned read causes 2 seeks • HBase checksums comes with 0.94 (HDP 1.2+). HBASE-5074. Page 26 Architecting the Future of Big Data blk_123456789 .blk_123456789.meta : Data chunk (dfs.bytes-per-checksum, 512 bytes) : Checksum chunk (4 bytes)
  • 27. © Hortonworks Inc. 2011 Checksums Page 27 Architecting the Future of Big Data • HFile version 2.1 writes checksums per Hfile block • HDFS checksum verification is bypassed on block read, will be done by HBase • If checksum fails, we go back to reading checksums from HDFS for “some time” • Due to double checksum bug(HDFS-3429) in remote reads in Hadoop 1, not enabled by default for now. Benchmark it yourself hbase.regionserver.checksum.verify = true hbase.hstore.bytes.per.checksum = 16384 hbase.hstore.checksum.algorithm = CRC32C Never set this: dfs.client.read.shortcircuit.skip.checksum = false HFile : Hfile data block chunk : Checksum chunk Hfile block : Block header
  • 28. © Hortonworks Inc. 2011 Rack 1 / Server 1 DataNode Default Block Placement Policy Page 28 Architecting the Future of Big Data b1 RegionServer Region A Region B StoreFile StoreFile StoreFile StoreFile StoreFile b2 b2 b9 b1 b1 b2 b3 b2 b1 b2b1 Rack N / Server M DataNode b2 b1 b1 Rack L / Server K DataNode b2 b1 Rack X / Server Y DataNode b1b2 b2 b3 RegionServer RegionServer RegionServer
  • 29. © Hortonworks Inc. 2011 Data locality for HBase • Poor data locality when the region is moved: – As a result of load balancing – Region server crash + failover • Most of the data won’t be local unless the files are compacted • Idea (from Facebook): Regions have affiliated nodes (primary, secondary, tertiary), HBASE-4755 • When writing a data file, give hints to the NN that we want these locations for block replicas (HDFS-2576) • LB should assign the region to one of the affiliated nodes on server crash – Keep data locality – SSR will still work • Reduces data loss probability Page 29 Architecting the Future of Big Data
  • 30. © Hortonworks Inc. 2011 Rack X / Server Y RegionServer Rack L / Server K RegionServer Rack N / Server M RegionServer Rack 1 / Server 1 Default Block Placement Policy Page 30 Architecting the Future of Big Data RegionServer Region A StoreFile StoreFile StoreFile Region B StoreFile StoreFile DataNode b1 b2 b2 b9 b1 b1 b2 b3 b2 b1 b2b1 DataNode b1 b2 b2 b9b1 b2 b1 DataNode b1 b2 b2 b9 b2 b1 DataNode b1 b2 b3 b2 b1
  • 31. © Hortonworks Inc. 2011 Other considerations • HBase riding over Namenode HA – Both Hadoop 1 (NFS based) and Hadoop 2 HA (JQM, etc) – Heavily tested with full stack HA • Retry HDFS operations • Isolate FileSystem usage from HBase internals • Hadoop 2 vs Hadoop 1 performance – Hadoop 2 is coming! • HDFS snapshots vs HBase snapshots – HBase DOES NOT use HDFS snapshots – Need hardlinks – Super flush API • HBase security vs HDFS security – All files are owned by HBase principal – No ACL’s in HDFS. Allowing a user to read Hfiles / snapshots directly is hard Page 31 Architecting the Future of Big Data
  • 32. © Hortonworks Inc. 2011 Open Topics • HDFS hard links – Rethink how we do snapshots, backups, etc • Parallel writes for WAL – Reduce latency on WAL syncs • SSD storage, cache – SSD storage type in Hadoop or local filesystem – Using SSD’s as a secondary cache – Selectively places tables / column families on SSD • HDFS zero-copy reads (HDFS-3051, HADOOP-8148) • HDFS inline checksums (HDFS-2699) • HDFS Quorum reads (HBASE-7509) Page 32 Architecting the Future of Big Data
  • 33. © Hortonworks Inc. 2011 Thanks Questions? Architecting the Future of Big Data Page 33 Enis Söztutar enis [ at ] apache [dot] org @enissoz