SQLFire at VMworld Europe 2011

Managing High Performance Data with vFabric
SQLFire

Carter Shanklin
Product Manager for vFabric

© 2009 VMware Inc. All rights reserved

Agenda

 What is SQLFire?
 Why SQL vs. NoSQL
 Why SQLFire versus other SQL databases
 SQLFire features + Demo
 How SQLFire Scales
• Hash partitioning
• Entity groups and collocation
• Data-aware stored procedures
 Consistency model
 Shared nothing persistence

What is vFabric SQLFire?

 SQLFire is a memory-optimized distributed SQL database.
 SQLFire attacks scalability challenges in two ways:
• Relaxes ACID semantics somewhat (in transactions and in replication)
• Horizontally scalable. Add capacity by adding nodes.
 SQLFire has built-in high availability and native support for
replication to multiple datacenters.
 SQLFire provides a real SQL interface. Ships with JDBC and
ADO.NET bindings with more to come.
 SQLFire can also be used as a cache in front of other databases.

SQLFire at-a-glance

Java C#
Client Client JDBC or ADO.NET
JDBC

Many physical machine nodes appear
as one logical system

As data changes,
subscribers are pushed Data transparently replicated and/or partitioned;
notification events Redundant storage can be in memory and/or on
disk

Increase/Decrease
capacity on the fly

Shared Nothing disk Synchronous read through,
persistence write through or
Each cache instance can
optionally persist to disk Asynchronous write-behind to
other data sources and sinks
File Other
system
Databases

4

The database world is changing.

 Many new data models (NoSQL) are emerging
• Key-value
• Column family (inspired by Google BigTable)
• Document
• Graph
 Most focus on making model less rigid than SQL
 Consistency model is not ACID

Low scale High scale Very high scale

STRICT – Tunable Eventual
Full ACID Consistency
(RDB)

 Different tradeoffs for different goals

SQLFire Versus NoSQL

Attribute NoSQL SQLFire
DB Interface Idiosyncratic (i.e. each is Standard SQL.
custom).
Querying Idiosyncratic or not present. SQL Queries.
Data Tunable, most favor eventual Tunable, favors high
Consistency consistency. consistency.
Transactions Weak or not present. Linearly scalable transaction
model.
Interface Design Designed for simplicity. Designed for compatibility.
Data Model Wide variety of different Relational model.
models.
Schema Focus on extreme flexibility, SQL model, requires DB
Flexibility dynamism. migrations, etc.

SQLFire Versus Other SQL Databases

Attribute SQLFire Other SQL DBs
DB Interface Standard SQL. Standard SQL.
Data Tunable. Mix of eventual High consistency.
Consistency consistency and high
consistency.
Transactions Supported. Very strong support.
Scaling Model Scale out, commodity Scale up.
servers.

SQLFire challenges traditional DB design, not SQL

Buffers primarily
tuned for IO
First write
to LOG

Second write to
Data files

 Too much I/O
 Design roots don‟t necessarily apply today
• Too much focus on ACID
8
• Disk synchronization bottlenecks
Confidential

SQLFire 1.0 Notable Features

 Horizontally scalable with Partitioning and Replication
 Multiple Topologies
• Client/Server, Asynchronous replication over WAN
 Queries
• Distributed and memory-optimized
 Procedures and Functions
• Standard Java stored procedures with “data awareness”
 Caching
• Loader, writers, Eviction, Overflow and Expiration
 Event framework
• Listeners, triggers, Asynchronous write behind
 Command line tools
 Manageability, Security

Scaling SQLFire

Partitioning & Replication

How SQLFire scales a common DB schema.

FLIGHTAVAILABILITY
---------------------------------------------

FLIGHTS FLIGHT_ID CHAR(6) NOT NULL ,
SEGMENT_NUMBER INTEGER NOT NULL ,
---------------------------------------------
FLIGHT_DATE DATE NOT NULL ,
ECONOMY_SEATS_TAKEN INTEGER ,
FLIGHT_ID CHAR(6) NOT NULL ,
…..
ORIG_AIRPORT CHAR(3), 1–M
PRIMARY KEY ( FLIGHT_ID,
DEPART_TIME TIME,
SEGMENT_NUMBER,
…..
FLIGHT_DATE))
PRIMARY KEY (FLIGHT_ID,
FOREIGN KEY (FLIGHT_ID,
SEGMENT_NUMBER)
SEGMENT_NUMBER)
REFERENCES FLIGHTS (
FLIGHT_ID,
SEGMENT_NUMBER)
1–1

FLIGHTHISTORY
---------------------------------------------
SEVERAL CODE/DIMENSION TABLES
FLIGHT_ID CHAR(6), ---------------------------------------------
SEGMENT_NUMBER INTEGER,
ORIG_AIRPORT CHAR(3), AIRLINES: AIRLINE INFORMATION (VERY STATIC)
DEPART_TIME TIME, COUNTRIES : LIST OF COUNTRIES SERVED BY FLIGHTS
DEST_AIRPORT CHAR(3), CITIES:
….. MAPS: PHOTOS OF REGIONS SERVED

Assume, thousands of flight rows, millions of flightavailability records

Creating Tables

CREATE TABLE AIRLINES (
AIRLINE CHAR(2) NOT NULL PRIMARY KEY,
AIRLINE_FULL VARCHAR(24),
BASIC_RATE DOUBLE PRECISION,
DISTANCE_DISCOUNT DOUBLE PRECISION,…. );

Table

SQLF SQLF SQLF

Replicated Tables

CREATE TABLE AIRLINES (
AIRLINE CHAR(2) NOT NULL PRIMARY KEY,
AIRLINE_FULL VARCHAR(24),
BASIC_RATE DOUBLE PRECISION,
DISTANCE_DISCOUNT DOUBLE PRECISION,…. )
REPLICATE;

Replicated Table Replicated Table Replicated Table

SQLF SQLF SQLF

Partitioned Tables

CREATE TABLE FLIGHTS (
ORIG_AIRPORT CHAR(3),
DEST_AIRPORT CHAR(3)
DEPART_TIME TIME,
FLIGHT_MILES INTEGER NOT NULL)
PARTITION BY COLUMN(FLIGHT_ID);

Replicated Table Replicated Table
Table Replicated Table
Partitioned Table Partitioned Table Partitioned Table

SQLF SQLF SQLF

Partition Redundancy

CREATE TABLE FLIGHTS (
ORIG_AIRPORT CHAR(3),
DEST_AIRPORT CHAR(3)
DEPART_TIME TIME,
FLIGHT_MILES INTEGER NOT NULL)
PARTITION BY COLUMN (FLIGHT_ID) REDUNDANCY 1;


Redundant Partition Redundant Partition Redundant Partition

SQLF SQLF SQLF

Partition Colocation

CREATE TABLE FLIGHTAVAILABILITY (
ECONOMY_SEATS_TAKEN INTEGER DEFAULT 0, …)
PARTITION BY COLUMN (FLIGHT_ID)
COLOCATE WITH (FLIGHTS);

Colocated Partition Colocated Partition Colocated Partition

SQLF SQLF SQLF

Persistent Tables

CREATE TABLE FLIGHTAVAILABILITY (
ECONOMY_SEATS_TAKEN INTEGER DEFAULT 0, …)
PARTITION BY COLUMN (FLIGHT_ID) COLOCATE WITH (FLIGHTS)
PERSISTENT persistentStore ASYNCHRONOUS;

Data dictionary is always persisted in each server

sqlf backup /export/fileServerDirectory/sqlfireBackupLocation

Colocated Partition Colocated Partition Colocated Partition

SQLF SQLF SQLF

Demo
Scaling with partitioned tables.

FLIGHTAVAILABILITY
---------------------------------------------

FLIGHTS FLIGHT_ID CHAR(6) NOT NULL ,
---------------------------------------------
ECONOMY_SEATS_TAKEN INTEGER ,
…..
ORIG_AIRPORT CHAR(3), 1–M
PRIMARY KEY ( FLIGHT_ID,
DEPART_TIME TIME,
SEGMENT_NUMBER,
…..
FLIGHT_DATE))
PRIMARY KEY (FLIGHT_ID,
FOREIGN KEY (FLIGHT_ID,
SEGMENT_NUMBER)
SEGMENT_NUMBER)
REFERENCES FLIGHTS (
FLIGHT_ID,
SEGMENT_NUMBER)
1–1

FLIGHTHISTORY
---------------------------------------------
SEVERAL CODE/DIMENSION TABLES
FLIGHT_ID CHAR(6), ---------------------------------------------
SEGMENT_NUMBER INTEGER,
ORIG_AIRPORT CHAR(3), AIRLINES: AIRLINE INFORMATION (VERY STATIC)
DEPART_TIME TIME, COUNTRIES : LIST OF COUNTRIES SERVED BY FLIGHTS
DEST_AIRPORT CHAR(3), CITIES:
….. MAPS: PHOTOS OF REGIONS SERVED

Hash partitioning for linear scaling

Key Hashing provides single hop access to its partition
But, what if the access is not based on the key … say, joins are involved

Pure hash-based partitioning will only get you so far.

 Consider this query :
select * from flights, flightAvailability
where flights.id = flightAvailability.flightid
and flight.fromAirport = ‘CPH’;

 If both tables are simply hash partitioned the join logic will
need execution on all nodes where flightavailability data is
stored.

 This will not scale.
• joins across distributed nodes could involve distributed locks and
potentially a lot of intermediate data transfer across nodes

To scale we need partition-aware DB design.

 DB architect must think about how data maps to partitions.
 The main idea is to:
• minimize excessive data distribution by keeping the most frequently accessed
and joined data collocated on partitions.
 Read Pat Helland‟s “Life beyond Distributed Transactions” and the
Google MegaStore paper.

SQLFire allows partition-aware design with the “colocated” keyword.

Entity Groups

FlightID is the
entity group Key

Table FlightAvailability partitioned by FlightID colocated with Flights

Solving this scalability problem with SQLFire.

 Create flightAvailability as follows:
CREATE TABLE flightAvailability
…
partitioned by flightid colocate with flights;

 Re-run the query:
select * from flights, flightAvailability
where flights.id = flightAvailability.flightid
and flight.fromAirport = ‘CPH’;

 The query is restricted to nodes containing flights with CPH
as the fromAirport.

More about partition-aware database design.

 OLTP systems tend to be partitionable.
• Typically it is the number of entities that grows over time and not the size of
the entity.
 Customer count perpetually grows, not the size of the customer info

• Most often access is very restricted and based on select entities
 given a FlightID, fetch flightAvailability records
 given a customerID, add/remove orders, shipment records

 Identify partition key for “Entity Group”
• "entity groups": set of entities across several related tables that can all share a
single identifier
 flightID is shared between the parent and child tables
 CustomerID shared between customer, order and shipment tables

Scaling Application logic with
Parallel “Data Aware procedures”

Stored Procedures in SQLFire.

 SQLFire stored procedures.
• Written in pure Java rather than proprietary extensions.
• Created and defined based on SQL standards.
• Supports “data awareness” and run only on nodes where applicable data
resides.
• They support a map/reduce-like execution style.
 Benefits:
• Write procedures in pure Java or take advantage of existing Java libraries.
• Easily take advantage of SQLFire as a highly scalable distributed system.

Procedures

Java Stored Procedures may be created according to the
SQL Standard

CREATE PROCEDURE getOverBookedFlights
(IN argument OBJECT, OUT result OBJECT)
LANGUAGE JAVA PARAMETER STYLE JAVA
READS SQL DATA DYNAMIC RESULT SETS 1
EXTERNAL NAME com.acme.OverBookedFLights;

SQLFire also supports the JDBC type Types.JAVA_OBJECT. A parameter of type
JAVA_OBJECT supports an arbitrary Serializable Java object.

In this case, the procedure will be executed on the server to
which a client is connected (or locally for Peer Clients)

Data Aware Procedures

Parallelize procedure and prune to nodes with required data

Extend the procedure call with the following syntax:
Client
CALL [PROCEDURE]
procedure_name
( [ expression [, expression ]* ] )
[ WITH RESULT PROCESSOR processor_name ]
[ { ON TABLE table_name [ WHERE whereClause ] }
|
{ ON {ALL | SERVER GROUPS
(server_group_name [, server_group_name ]*) }} Fabric Server 1 Fabric Server 2

]

CALL getOverBookedFlights( <bind arguments>
ON TABLE FLIGHTAVAILABILITY
Hint the data the procedure depends on
WHERE FLIGHTID = <SomeFLIGHTID> ;

If table is partitioned by columns in the where clause the procedure execution is pruned to nodes with
the data (node with <someFLIGHTID> in this case)

Parallelize procedure then aggregate (reduce)

CALL [PROCEDURE]
procedure_name
register a Java Result Processor (optional in some cases):
( [ expression [, expression ]* ] )
CALL SQLF.CreateResultProcessor( [ WITH RESULT PROCESSOR processor_name ]
processor_name,
processor_class_name); [ { ON TABLE table_name [ WHERE whereClause ] } |
{ ON {ALL | SERVER GROUPS
(server_group_name [, server_group_name ]*) }}
]

Client

Fabric Server 1 Fabric Server 2 Fabric Server 3

Consistency Model without Transactions

 Replication within cluster is always eager and synchronous
 Row updates are always atomic; No need to use transactions
 FIFO consistency: writes performed by a single thread are seen by
all other processes in the order in which they were issued
 Consistency in Partitioned tables
• a partitioned table row owned by one member at a point in time
• all updates are serialized to replicas through owner
• "Total ordering" at a row level: atomic and isolated
 Membership changes and consistency – need another hour 
 Pessimistic concurrency support using „Select for update‟
 Support for referential integrity

SQLFire Transactions

 Highly scalable without any centralized coordinator or lock
manager
 We make some important assumptions
• Most OLTP transactions are small in duration and size
• Write-write conflicts are very rare in practice
 How does it work?
• Each data node has a sub-coordinator to track TX state
• Eagerly acquire local “write” locks on each replica
 Object owned by a single primary at a point in time
• Fail fast if lock cannot be obtained
 Atomic and works with the cluster Failure detection system
 Isolated until commit
• Only support local isolation during commit

Scaling disk access with shared
nothing disk files and a
“journaling” store design

Disk persistence in SQLF

Memory Memory
Tables Tables

LOG LOG
Compressor Compressor

OS Buffers OS Buffers

Record1 Record1
Record1

Record2
Record2 Append only Record1

Record2
Record2 Append only
Record3
Record3
Operation logs Record3
Record3
Operation logs

 Parallel log structured
storage • Don’t seek to disk
 Each partition writes in • Don’t flush all the way to disk
parallel – Use OS scheduler to time
write
 Backups write to disk also
• Do this on primary + secondary
• Increase reliability against h/w
• Realize very high throughput
loss

How does it perform? Scale?

 Scale from 2 to 10 servers (one per host)
 Scale from 200 to 1200 simulated clients (10 hosts)
 Single partitioned table: int PK, 40 fields (20 ints, 20 strings)

How does it perform? Scale?

 CPU% remained low per server – about 30% indicating many more
clients could be handled

Is latency low with scale?

 Latency decreases with server capacity
 50-70% take < 1 millisecond
 About 90% take less than 2 milliseconds

SQLFire beta available now

http://vmware.com/go/sqlfire

Q&A

SQLFire at VMworld Europe 2011

More Related Content

What's hot

Similar to SQLFire at VMworld Europe 2011

Recently uploaded

SQLFire at VMworld Europe 2011

Editor's Notes