Emergent Distributed Data Storage

http://www.coordguru.com

Emergent Distributed Data Storages
for Big Data, Storage, and Analysis

Woohyun Kim
The creator of open source “Coord”
(http://www.coordguru.com)

2010-01-27

Contents

The Advent of Big Data MapReduce Debates
• Noah’s Ark Problem • MapReduce is just A Major Step Backwards!!!
• Key Issues with ‚Big Data‛ • RDB experts Jump the MR Shark
• How to deal with ‚Big Data‛ • DBs are hammers; MR is a screwdriver
• MR is a Step Backwards, but some Steps Forward

Hadoop Revolution A Hybrid of MapReduce and RDBMS
• Integrate MapReduce into RDBMS
• Best Practice in Hadoop
• In-Database MapReduce vs. File-only MapReduce
• Hadoop is changing the Game
• Big Data goes well with Hadoop Non-Relational Data Storages
• Case Study: Parallel Join • Throw ‘Relational’ Away, and Take ‘Schema-Free’
• Case Study: Further Study in Parallel Join • A Comparison of Non-Relational Data Storages
• Case Study: Improvements in Parallel Join • Emergent Document-oriented Storages
• Document-oriented vs. RDBMS


The Advent of Big Data


Noah’s Ark Problem
• Did Noah take dinosaurs on the Ark?
• The Ark was a very large ship designed especially for its important purpose
• It was so large and complex that it took Noah 120 years to build

• How to put such a big thing
• Diet?
• Split?
• Differentiate
• Put
• Integrate
• Scale Up?
• Scale Out?
• ‚Big Data‛ problem is just like that


Key Issues with ‚Big Data‛
• Lookup
• Metadata server -> centralized or distributed -> partitioned replicas to avoid a single of the failure

• Partition
• Data locality -> network bandwidth reduction -> putting the computation near the data

• Replication
• Hardwar Failure -> Data Loss -> Availability from redundant copies of the data

• Load-balanced Parallel Processing
• Corrupt Data or Remote process failure -> speculative execution or rescheduling

• Ad-hoc Analysis
• Some partitioned data may need to be combined with another data


How to deal with ‚Big Data‛
Struggling to STORE and ANALYZE ‚Big Data‛


Appendix: What is ETL?
ETL(Extract, Transform, and Load)
• A process in database usage and especially in data warehousing that involves:
• Extracting data from outside sources(such as different data
organization/format, non-relational database structures)
• Transforming it to fit operational needs (which can include quality levels)
• Selection, translation, encoding, calculation, filtering, sorting, joining,
aggregation, transposing or pivoting, splitting, disaggregation
• Loading it into the end target (database or data warehouse)

ETL Open Sources
• Talend Open Studio
• Pentaho Data Integration (Kettle)
• RapidMiner
• Jitterbit 2.0,
• Apatar
• Clover.ETL
• Scriptelle


Hadoop Revolution


Best Practice in Hadoop
• Software Stack in Google/Hadoop • Cookbook for ‚Big Data‛

• Structured Data Storage for ‚Big Data‛
Row key Column key

Row
Structured
Data

Time
Column Column stamp
Family Family


Appendix: What is MapReduce?
Map
• Read a set of ‚records’ from an input file, which acts as filtering or transformations
• Output a set of (key, data) pair, which partitions them into R disjoint buckets by
the key
Reduce
• Read a set of (key, a list of data) pairs from R disjoint buckets
• Each R from map’s outputs is shuffled, and aggregated into its corresponding
reduce with being ordering by the key
• Output a new set of records

Map
Reduce
Map
Reduce

Map

Group-By/ Aggregate/
Filter Aggregator


Hadoop is changing the Game
• Hadoop, DW, and BI


Big Data goes well with Hadoop
• Parallelize Relational Algebra Operations using MapReduce


Case Study: Parallel Join
• A Parallel Join Example using MapReduce


Case Study: Further Study in Parallel Join
Problems
• Need to sort
• Move the partitioned data across the network
• Due to shuffling, must send the whole data
• Skewed by popular keys
• All records for a particular key are sent to the same reducer
• Overhead by tagging

Alternatives
• Map-side Join
• Mapper-only job to avoid sort and to reduce data movement across the
network
• Semi-Join
• Shrink data size through semi-join(by preprocessing)


Case Study: Improvements in Parallel Join
Map-Side Join
• Replicate a relatively smaller input source to the cluster
• Put the replicated dataset into a local hash table
• Join – a relatively larger input source with each local hash table
• Mapper: do Mapper-side Join

Semi-Join
• Extract – unique IDs referenced in a larger input source(A)
• Mapper: extract Movie IDs from Ratings records
• Reducer: accumulate all unique Movie IDs
• Filter – the other larger input source(B) with the referenced unique IDs
• Mapper: filter the referenced Movie IDs from full Movie dataset
• Join - a larger input source(A) with the filtered datasets
• Mapper: do Mapper-side Join
• Ratings records & the filtered movie IDs dataset


MapReduce Debates


MapReduce is just A Major Step Backwards!!!
Dewitt and StoneBraker in January 17, 2008

• A giant step backward in the programming paradigm for
large-scale data intensive applications
• Schema are good
• Type check in runtime, so no garbage
• Separation of the schema from the application is good
• Schema is stored in catalogs, so can be queried(in SQL)
• High-level access languages are good
• Present what you want rather than an algorithm for how to get it
• No schema??!
• At least one data field by specifying the key as input
• For Bigtable/Hbase, different tuples within the same table can
actually have different schemas
• Even there is no support for logical schema changes such as
views


MapReduce is just A Major Step Backwards!!! (cont’d)

• A sub-optimal implementation, in that it uses brute force instead of
indexing
• Indexing
• All modern DBMSs use hash or B-tree indexes to accelerate access to data
• In addition, there is a query optimizer to decide whether to use an index or
perform a brute-force sequential search
• However, MapReduce has no indexes, so processes only in brute force fashion
• Automatic parallel execution
• In the 1980s, DBMS research community explored it such as Gamma, Bubba,
Grace, even commercial Teradata
• Skew
• The distribution of records with the same key causes is skewed in the map
phase, so it causes some reduce to take much longer than others
• Intermediate data pulling
• In the reduce phase, two or more reduce attempt to read input files form the
same map node simultaneously



• Not novel at all – it represents a specific implementation of well
known techniques developed nearly 25 years ago
• Partitioning for join
• Application of Hash to Data Base Machine and its Architecture, 1983
• Joins in parallel on a shared-nothing
• Multiprocessor Hash-based Join Algorithms, 1985
• The Case for Shared-Nothing, 1986
• Aggregates in parallel
• The Gamma Database Machine Project, 1990
• Parallel Database System: The Future of High Performance Database Systems,
1992
• Adaptive Parallel Aggregation Algorithms, 1995
• Teradata has been selling a commercial DBMS utilizing all of these
techniques for more than 20 years
• PostgreSQL supported user-defined functions and user-defined
aggregates in the mid 1980s



• Missing most of the features that are routinely included in current DBMS
• MapReduce provides only a sliver of the functionality found in modern DBMSs
• Bulk loader – transform input data in files into a desired format and load it into a DBMS
• Indexing – hash or B-Tree indexes
• Updates – change the data in the data base
• Transactions – support parallel update and recovery from failures during update
• integrity constraints – help keep garbage out of the data base
• referential integrity – again, help keep garbage out of the data base
• Views – so the schema can change without having to rewrite the application program

• Incompatible with all of the tools DBMS users have come to depend on
• MapReduce cannot use the tools available in a modern SQL DBMS, and has none of
its own
• Report writers(Crystal reports)
• Prepare reports for human visualization
• business intelligence tools(Business Objects or Cognos)
• Enable ad-hoc querying of large data warehouses
• data mining tools(Oracle Data Mining or IBM DB2 Intelligent Miner)
• Allow a user to discover structure in large data sets
• replication tools(Golden Gate)
• Allow a user to replicate data from on DBMS to another
• database design tools(Embarcadero)
• Assist the user in constructing a data base


What the !@# MapReduce?


RDB experts Jump the MR Shark
Greg Jorgensen in January 17, 2008

• Arg1: MapReduce is a step backwards in database access
• MapReduce is not a database, a data storage, or management system
• MapReduce is an algorithmic technique for the distributed processing of large
amounts of data
• Arg2: MapReduce is a poor implementation
• MapReduce is one way to generate indexes from a large volume of data, but it’s not
a data storage and retrieval system
• Arg3: MapReduce is not novel
• Hashing, parallel processing, data partitioning, and user-defined functions are all old
hat in the RDBMS world, but so what?
• The big innovation MapReduce enables is distributing data processing across a
network of cheap and possibly unreliable computers
• Arg4: MapReduce is missing features
• Arg5: MapReduce is incompatible with the DBMS tools
• The ability to process a huge volume of data quickly such as web crawling and log
analysis is more important than guaranteeing 100% data integrity and completeness


DBs are hammers; MR is a screwdriver
Mark C. Chu-Carroll

• RDBs don’t parallelize very well
• How many RDBs do you know that can efficiently split a
task among 1,000 cheap computers?
• RDBs don’t handle non-tabular data well
• RDBs are notorious for doing a poor job on recursive data
structures
• MapReduce isn’t intended to replace relational
databases
• It’s intended to provide a lightweight way of programming
things so that they can run fast by running in parallel on a
lot of machines


MR is a Step Backwards, but some Steps Forward
Eugene Shekita

• Arg1: Data Models, Schemas, and Query Languages
• Semi-structured data model and high level of parallel data flow query language is
built on top of MapReduce
• Pig, Hive, Jaql, Cascading, Cloudbase
• Hadoop will eventually have a real data model, schema, catalogs, and query
language
• Moreover, Pig, Jaql, and Cascading are some steps forward
• Support semi-structured data
• Support more high level-like parallel data flow languages than declarative query
languages
• Greenplum and Aster Data support MapReduce, but look more limited than Pig, Jaql,
Cascading
• The calls to MapReduce functions wrapped in SQL queries will make it difficult
to work with semi-structured data and program multi-step dataflows
• Arg3: Novelty
• Teradata was doing parallel group-by 20 years ago
• UDAs and UDFs appeared in PostgreSQL in the mid 80s
• And yet, MapReduce is much more flexible, and fault-tolerant
• Support semi-structured data types, customizable partitioning


Lessons Learned from the Debates
Who Moved My Cheese?
• Speed
• The seek times of physical storage is not keeping pace with improvements
in network speeds
• Scale
• The difficulty of scaling the RDBMS out efficiently
• Clustering beyond a handful of servers is notoriously hard
• Integration
• Today’s data processing tasks increasingly have to access and combine
data from many different non-relational sources, often over a network
• Volume
• Data volumes have grown from tens of gigabytes in the 1990s to
hundreds of terabytes and often petabytes in recent years

Stolen from 10 Ways To complement the Enterprise RDBMS using Hadoop


A Hybrid of MapReduce and RDBMS


Integrate MapReduce into RDBMS
RDBMS MapReduce
Data size Gigabytes Petabytes
Updates Read and write(Mutable) Write once, read many times(Immutable)
Latency Low High
Access Interactive(point query) and batch Batch(ad-hoc query in brute-force)
Structure Fixed schema Semi-structured schema
Language SQL Procedural (Java, C++, etc)
Integrity High Low
Scaling Nonlinear Linear

HadoopDB Greenplum Aster Data


In-Database MapReduce vs. File-only MapReduce
• In-Database MapReduce
• Greenplum, Aster Data, HadoopDB
• File-only MapReduce
• Pig, Hive, Cloudbase

In-Database MapReduce File-Only MapReduce
Target User Analyst, DBA, Data Miner Computer Science Engineer
Scale & Performance High High
Hardware Costs Low Low
Analytical Insights High High
Failover & Recovery High High
Use: Ad-Hoc Queries Easy (seamless) Harder (custom)
Use: UI, Client Tools BI Tool (GUI), SQL (CLI) Developer Tool (Java)
Use: Ecosystem High (JDBC, ODBC) Lower (custom)
Protect: Data Integrity High (ACID, schema) Lower (no transaction guarantees)
Protect: Security High (roles, privileges) Lower (custom)
Protect: Backup & DR High (database backup/DR) Lower (custom)
Performance: Mixed Workloads High (workload/QoS mgmt) Lower (limited concurrency)
Performance: Network Bottleneck No (optimized partitioning) Higher (network inefficient)
Operational Cost Low (1 DBA) Higher (several engineers)


Non-Relational Data Storages


Throw ‘Relational’ Away, and Take ‘Schema-Free’
The new face of data
• Scale out, not up
• Online load balancing, cluster growth
• Flexible schema
• Some data have sparse attributes, do not need ‘relational’ property
• Document/Term vector, User/Item matrix, Log-structured data
• Key-oriented queries
• Some data are stored and retrieved mainly by primary key, without complex joins
• Trade-off of Consistency, Availability, and Partition Tolerance

Two of Feasible Approaches
• Bigtable
• How can we build a distributed DB on top of GFS?
• B+ Tree style Lookup, Synchronized consistency
• Memtable/Commit Log/Immutable SSTable/Indexes, Compaction

• Dynamo
• How can we build a distributed hash table appropriate for the data center?
• DHT style Lookup, Eventually consistency


A Comparison of Non-Relational Data Storages
Name Language Fault-tolerance Persistence Client Protocol Data model Docs Community
Hbase Java Replication, partitioning Custom on-disk Custom API, Thrift, Rest Bigtable A Apache, yes
Zvents, Baidu, y
Hypertable C++ Replication, partitioning Custom on-disk Thrift, other Bigtable A
es
Neptune Java Replication, partitioning Custom on-disk Custom API, Thrift, Rest Bigtable A NHN, some
partitioned, replicated, read-rep Pluggable: BerkleyDB, Mys Structured /
Voldemort Java Java API A Linkedin, no
air ql blob / text
partitioned, replicated, immutab Custom on-disk (append o
Ringo Erlang HTTP blob B Nokia, no
le nly log)
Scalaris Erlang partitioned, replicated, paxos In-memory only Erlang, Java, HTTP blob B OnScale, no
Kai Erlang partitioned, replicated? On-disk Dets file Memcached blob C no
Dynomite Erlang partitioned, replicated Pluggable: couch, dets Custom ascii, Thrift blob D+ Powerset, no
MemcacheDB C replication BerkleyDB Memcached blob B some
Pluggable: BerkleyDB, Cust Document or Third rail, unsur
ThruDB C++ Replication Thrift C+
om, Mysql, S3 iented e
Document or
CouchDB Erlang Replication, partitioning? Custom on-disk HTTP, json A Apache, yes
iented (json)
Bigtable me
Cassandra Java Replication, partitioning Custom on-disk Thrift F Facebook, no
ets Dynamo
Pluggable: in-memory, Luc
Coord C++ Replication?, partitioning Custom API, Thrift text / blob A NHN, some
ene, BerkelyDB, Mysql
Stolen from Anti-RDBMS - A list of distributed key-value stores by Richard Jones
Bigtable DHT

HBase Cassandra Dynamo
Hypertable Voldemort
Dynomite KAI
SimpleDB
Chordless CouchDB
Tokyo Cabinet MongoDB
MemcacheDB
ThruDB
Scalaris
Document-oriented
Key-Value On-going classification by Woohyun Kim


Emergent Document-oriented Storages
Why Document-oriented?
• All fields become optional
• All relationships become Many-to-Many
• Chatter always expands

Key Features
• Schema-Free
• Straightforward Data Model
• Full Text Indexing
• RESTful HTTP/JSON API


Document-oriented vs. RDBMS
CouchDB MongoDB MySQL
Data Model Document-Oriented (JSON) Document-Oriented (BSON) Relational
string, int, double, boolean, date, bytea
Data Types ? Link
rray, object, array, others
Large Objects (Files) Yes (attachments) Yes (GridFS) no???
Master-master (with developer sup
Replication Master-slave Master-slave
plied conflict resolution)
Object(row) Storage One large repository Collection based Table based
Map/reduce of javascript functions Dynamic; object-based query language
Query Method Dynamic; SQL
to lazily build an index per query
Secondary Indexes Yes Yes Yes
Atomicity Single document Single document Yes – advanced
Interface REST Native drivers Native drivers
Server-side batch dat
? Yes, via javascript Yes (SQL)
a manipulation
Written in Erlang C++ C
Concurrency Control MVCC Update in Place Update in Place


Thank you.


Appendix: What is Coord?
Architectural Comparison
• dust: a distributed file system based on DHT
• coord spaces: a resource sharable store system based on SBA
• coord mapreduce: a simplified large-scale data processing framework
• warp: a scalable remote/parallel execution system
• graph: a large-scale distributed graph search system


Appendix: Coord Internals
 A space-based architecture built on distributed hash tables
 SBA(Space-based Architecture)
 processes communicate with others thru. only spaces
 DHT(Distributed Hash Tables)
 data identified by hash functions are placed on numerically near nodes
 A computing platform to project a single address space on
distributed memories
 As if users worked in a single computing environment

App

take write
read
2m-1 0

node 1 node 2 node 3 node n

Emergent Distributed Data Storage

More Related Content

What's hot (20)

Viewers also liked (14)

Similar to Emergent Distributed Data Storage (20)

Recently uploaded (20)

Emergent Distributed Data Storage