DEVELOPING POLYGLOT PERSISTENCE APPLICATIONS
Chris RichardsonAuthor of POJOs in ActionFounder of the original CloudFoundry.com @crichardson [email protected]://plainoldobjects.com
@crichardson
Presentation goal
The benefits and drawbacks of polyglot persistence
and
How to design applications that use this approach
@crichardson
About Chris
@crichardson
(About Chris)
@crichardson
About Chris()
@crichardson
About Chris
@crichardson
About Chris
http://www.theregister.co.uk/2009/08/19/springsource_cloud_foundry/
@crichardson
vmc push About-Chris
Developer Advocate for CloudFoundry.com
Signup at http://cloudfoundry.com
@crichardson
Agenda
• Why polyglot persistence?
• Using Redis as a cache
• Optimizing queries using Redis materialized views
• Synchronizing MySQL and Redis
• Tracking changes to entities
• Using a modular asynchronous architecture
@crichardson
Food to Go
• Take-out food delivery service
• “Launched” in 2006
@crichardson
Food To Go Architecture
Order taking
Restaurant Management
MySQL Database
CONSUMER RESTAURANT OWNER
@crichardson
Success ⇒ Growth challenges
• Increasing traffic
• Increasing data volume
• Distribute across a few data centers
• Increasing domain model complexity
@crichardson
Limitations of relational databases
• Scalability
• Distribution
• Schema updates
• O/R impedance mismatch
• Handling semi-structured data
@crichardson
Solution: Spend Money
http://upload.wikimedia.org/wikipedia/commons/e/e5/Rising_Sun_Yacht.JPG
OR
http://www.trekbikes.com/us/en/bikes/road/race_performance/madone_5_series/madone_5_2/#
@crichardson
Solution: Use NoSQL
Benefits
• Higher performance
• Higher scalability
• Richer data-model
• Schema-less
Drawbacks
• Limited transactions
• Limited querying
• Relaxed consistency
• Unconstrained data
@crichardson
Example NoSQL Databases
Database Key features
Cassandra Extensible column store, very scalable, distributed
Neo4j Graph database
MongoDB Document-oriented, fast, scalable
Redis Key-value store, very fast
http://nosql-database.org/ lists 122+ NoSQL databases
@crichardson
• Advanced key-value store
• C-based server
• Very fast, e.g. 100K reqs/sec
• Optional persistence
• Transactions with optimistic locking
• Master-slave replication
• Sharding using client-side consistent hashing
RedisK1 V1
K2 V2
... ...
@crichardson
Sorted sets
myset a
5.0
b
10.
Key Value
ScoreMembers are sorted by score
@crichardson
Adding members to a sorted setRedis Server
zadd myset 5.0 a myset a
5.0
Key Score Value
@crichardson
Adding members to a sorted setRedis Server
zadd myset 10.0 b myset a
5.0
b
10.
@crichardson
Adding members to a sorted setRedis Server
zadd myset 1.0 c myset a
5.0
b
10.
c
1.0
@crichardson
Retrieving members by index range
Redis Server
zrange myset 0 1
myset a
5.0
b
10.
c
1.0ac
Key StartIndex
EndIndex
@crichardson
Retrieving members by score
Redis Server
zrangebyscore myset 1 6
myset a
5.0
b
10.
c
1.0ac
Key Min value
Max value
@crichardson
Redis use cases• Replacement for Memcached
• Session state
• Cache of data retrieved from system of record (SOR)
• Replica of SOR for queries needing high-performance
• Handling tasks that overload an RDBMS
• Hit counts - INCR
• Most recent N items - LPUSH and LTRIM
• Randomly selecting an item – SRANDMEMBER
• Queuing – Lists with LPOP, RPUSH, ….
• High score tables – Sorted sets and ZINCRBY
• …
@crichardson
Redis is great but there are tradeoffs
• Low-level query language: PK-based access only
• Limited transaction model:
• Read first and then execute updates as batch
• Difficult to compose code
• Data must fit in memory
• Single-threaded server : run multiple with client-side sharding
• Missing features such as access control, ...
@crichardson
And don’t forget:
An RDBMS is fine for many applications
@crichardson
The future is polyglot
IEEE Software Sept/October 2010 - Debasish Ghosh / Twitter @debasishg
e.g. Netflix• RDBMS• SimpleDB• Cassandra• Hadoop/Hbase
@crichardson
Agenda
• Why polyglot persistence?
• Using Redis as a cache
• Optimizing queries using Redis materialized views
• Synchronizing MySQL and Redis
• Tracking changes to entities
• Using a modular asynchronous architecture
@crichardson
Food to Go – Domain model (partial)
class Restaurant { long id; String name; Set<String> serviceArea; Set<TimeRange> openingHours; List<MenuItem> menuItems;}
class MenuItem { String name; double price;}
class TimeRange { long id; int dayOfWeek; int openTime; int closeTime;}
@crichardson
Database schemaID Name …
1 Ajanta
2 Montclair Eggshop
Restaurant_id zipcode
1 94707
1 94619
2 94611
2 94619
Restaurant_id dayOfWeek openTime closeTime
1 Monday 1130 1430
1 Monday 1730 2130
2 Tuesday 1130 …
RESTAURANT table
RESTAURANT_ZIPCODE table
RESTAURANT_TIME_RANGE table
@crichardson
RestaurantRepository
public interface RestaurantRepository { void addRestaurant(Restaurant restaurant); Restaurant findById(long id); ...}
Food To Go will have scaling eventually issues
@crichardson
Increase scalability by caching
Order taking
Restaurant Management
MySQL Database
CONSUMER RESTAURANT OWNER
Cache
@crichardson
Caching Options
• Where:
• Hibernate 2nd level cache
• Explicit calls from application code
• Caching aspect
• Cache technologies: Ehcache, Memcached, Infinispan, ...
Redis is also an option
@crichardson
Using Redis as a cache
• Spring 3.1 cache abstraction
• Annotations specify which methods to cache
• CacheManager - pluggable back-end cache
• Spring Data for Redis
• Simplifies the development of Redis applications
• Provides RedisTemplate (analogous to JdbcTemplate)
• Provides RedisCacheManager
@crichardson
Using Spring 3.1 Caching@Servicepublic class RestaurantManagementServiceImpl implements RestaurantManagementService {
private final RestaurantRepository restaurantRepository;
@Autowired public RestaurantManagementServiceImpl(RestaurantRepository restaurantRepository) { this.restaurantRepository = restaurantRepository; } @Override public void add(Restaurant restaurant) { restaurantRepository.add(restaurant); }
@Override @Cacheable(value = "Restaurant") public Restaurant findById(int id) { return restaurantRepository.findRestaurant(id); }
@Override @CacheEvict(value = "Restaurant", key="#restaurant.id") public void update(Restaurant restaurant) { restaurantRepository.update(restaurant); }
Cache result
Evict from cache
@crichardson
Configuring the Redis Cache Manager
<cache:annotation-driven />
<bean id="cacheManager" class="org.springframework.data.redis.cache.RedisCacheManager" > <constructor-arg ref="restaurantTemplate"/> </bean>
Enables caching
Specifies CacheManager implementation
The RedisTemplate used to access Redis
@crichardson
TimeRange MenuItem
Domain object to key-value mapping?
Restaurant
TimeRange MenuItem
ServiceArea
K1 V1
K2 V2
... ...
@crichardson
RedisTemplate handles the mapping
• Principal API provided by Spring Data to Redis
• Analogous to JdbcTemplate
• Encapsulates boilerplate code, e.g. connection management
• Maps Java objects ⇔ Redis byte[]’s
@crichardson
Serializers: object ⇔ byte[]
• RedisTemplate has multiple serializers
• DefaultSerializer - defaults to JdkSerializationRedisSerializer
• KeySerializer
• ValueSerializer
• HashKeySerializer
• HashValueSerializer
@crichardson
Serializing a Restaurant as JSON@Configurationpublic class RestaurantManagementRedisConfiguration {
@Autowired private RestaurantObjectMapperFactory restaurantObjectMapperFactory; private JacksonJsonRedisSerializer<Restaurant> makeRestaurantJsonSerializer() { JacksonJsonRedisSerializer<Restaurant> serializer =
new JacksonJsonRedisSerializer<Restaurant>(Restaurant.class); ... return serializer; }
@Bean @Qualifier("Restaurant") public RedisTemplate<String, Restaurant> restaurantTemplate(RedisConnectionFactory factory) { RedisTemplate<String, Restaurant> template = new RedisTemplate<String, Restaurant>(); template.setConnectionFactory(factory); JacksonJsonRedisSerializer<Restaurant> jsonSerializer = makeRestaurantJsonSerializer(); template.setValueSerializer(jsonSerializer); return template; }
}
Serialize restaurants using Jackson JSON
@crichardson
Caching with Redis
Order taking
Restaurant Management
MySQL Database
CONSUMER RESTAURANT OWNER
RedisCacheFirst Second
@crichardson
Agenda
• Why polyglot persistence?
• Using Redis as a cache
• Optimizing queries using Redis materialized views
• Synchronizing MySQL and Redis
• Tracking changes to entities
• Using a modular asynchronous architecture
@crichardson
Finding available restaurants
Available restaurants =Serve the zip code of the delivery address
AND Are open at the delivery time
public interface AvailableRestaurantRepository {
List<AvailableRestaurant> findAvailableRestaurants(Address deliveryAddress, Date deliveryTime); ...}
@crichardson
Finding available restaurants on Monday, 6.15pm for 94619 zipcode
Straightforward three-way join
select r.*from restaurant r inner join restaurant_time_range tr on r.id =tr.restaurant_id inner join restaurant_zipcode sa on r.id = sa.restaurant_id where ’94619’ = sa.zip_code and tr.day_of_week=’monday’ and tr.openingtime <= 1815 and 1815 <= tr.closingtime
@crichardson
How to scale queries?
@crichardson
Option #1: Query caching
• [ZipCode, DeliveryTime] ⇨ list of available restaurants
BUT
• Long tail queries
• Update restaurant ⇨ Flush entire cache
Ineffective
@crichardson
Option #2: Master/Slave replication
MySQL Master
MySQL Slave 1
MySQL Slave 2
MySQL Slave N
Writes Consistent reads
Queries(Inconsistent reads)
@crichardson
Master/Slave replication
• Mostly straightforward
BUT
• Assumes that SQL query is efficient
• Complexity of administration of slaves
• Doesn’t scale writes
@crichardson
Option #3: Redis materialized views
Order taking
Restaurant Management
MySQL Database
CONSUMER RESTAURANT OWNER
CacheRedis
Systemof RecordCopy
findAvailable()update()
@crichardson
BUT how to implement findAvailableRestaurants() with Redis?!
?select r.*from restaurant r inner join restaurant_time_range tr on r.id =tr.restaurant_id inner join restaurant_zipcode sa on r.id = sa.restaurant_id where ’94619’ = sa.zip_code and tr.day_of_week=’monday’
and tr.openingtime <= 1815 and 1815 <= tr.closingtime
K1 V1
K2 V2
... ...
@crichardson
Solution: Build an index using sorted sets and ZRANGEBYSCORE
ZRANGEBYSCORE myset 1 6
select valuefrom sorted_setwhere key = ‘myset’ and score >= 1 and score <= 6
=
key value score
sorted_set
@crichardson
select r.*from restaurant r inner join restaurant_time_range tr on r.id =tr.restaurant_id inner join restaurant_zipcode sa on r.id = sa.restaurant_id where ’94619’ = sa.zip_code and tr.day_of_week=’monday’ and tr.openingtime <= 1815 and 1815 <= tr.closingtime
select valuefrom sorted_setwhere key = ? and score >= ? and score <= ?
?
How to transform the SELECT statement?
@crichardson
We need to denormalize
Think materialized view
@crichardson
Simplification #1: Denormalization
Restaurant_id Day_of_week Open_time Close_time Zip_code
1 Monday 1130 1430 94707
1 Monday 1130 1430 94619
1 Monday 1730 2130 94707
1 Monday 1730 2130 94619
2 Monday 0700 1430 94619
…
SELECT restaurant_id FROM time_range_zip_code WHERE day_of_week = ‘Monday’ AND zip_code = 94619 AND 1815 < close_time AND open_time < 1815
Simpler query: No joins Two = and two <
@crichardson
Simplification #2: Application filtering
SELECT restaurant_id, open_time FROM time_range_zip_code WHERE day_of_week = ‘Monday’ AND zip_code = 94619 AND 1815 < close_time AND open_time < 1815
Even simpler query• No joins• Two = and one <
@crichardson
Simplification #3: Eliminate multiple =’s with concatenation
SELECT restaurant_id, open_time FROM time_range_zip_code WHERE zip_code_day_of_week = ‘94619:Monday’ AND 1815 < close_time
Restaurant_id Zip_dow Open_time Close_time
1 94707:Monday 1130 1430
1 94619:Monday 1130 1430
1 94707:Monday 1730 2130
1 94619:Monday 1730 2130
2 94619:Monday 0700 1430
…
key
range
@crichardson
Simplification #4: Eliminate multiple RETURN VALUES with concatenation
SELECT open_time_restaurant_id, FROM time_range_zip_code WHERE zip_code_day_of_week = ‘94619:Monday’ AND 1815 < close_time
zip_dow open_time_restaurant_id close_time
94707:Monday 1130_1 1430
94619:Monday 1130_1 1430
94707:Monday 1730_1 2130
94619:Monday 1730_1 2130
94619:Monday 0700_2 1430
...
✔
@crichardson
zip_dow open_time_restaurant_id close_time
94707:Monday 1130_1 1430
94619:Monday 1130_1 1430
94707:Monday 1730_1 2130
94619:Monday 1730_1 2130
94619:Monday 0700_2 1430
...
Sorted Set [ Entry:Score, …]
[1130_1:1430, 1730_1:2130]
[0700_2:1430, 1130_1:1430, 1730_1:2130]
Using a Redis sorted set as an index
Key
94619:Monday
94707:Monday
94619:Monday 0700_2 1430
@crichardson
Querying with ZRANGEBYSCORE
ZRANGEBYSCORE 94619:Monday 1815 2359
{1730_1}
1730 is before 1815 Ajanta is open
Delivery timeDelivery zip and day
Sorted Set [ Entry:Score, …]
[1130_1:1430, 1730_1:2130]
[0700_2:1430, 1130_1:1430, 1730_1:2130]
Key
94619:Monday
94707:Monday
@crichardson
Adding a Restaurant
Text
@Componentpublic class AvailableRestaurantRepositoryImpl implements AvailableRestaurantRepository {
@Override public void add(Restaurant restaurant) { addRestaurantDetails(restaurant); addAvailabilityIndexEntries(restaurant); }
private void addRestaurantDetails(Restaurant restaurant) { restaurantTemplate.opsForValue().set(keyFormatter.key(restaurant.getId()), restaurant); } private void addAvailabilityIndexEntries(Restaurant restaurant) { for (TimeRange tr : restaurant.getOpeningHours()) { String indexValue = formatTrId(restaurant, tr); int dayOfWeek = tr.getDayOfWeek(); int closingTime = tr.getClosingTime(); for (String zipCode : restaurant.getServiceArea()) { redisTemplate.opsForZSet().add(closingTimesKey(zipCode, dayOfWeek), indexValue,
closingTime); } } }
key member
score
Store as JSON
@crichardson
Finding available Restaurants@Componentpublic class AvailableRestaurantRepositoryImpl implements AvailableRestaurantRepository { @Override public List<AvailableRestaurant>
findAvailableRestaurants(Address deliveryAddress, Date deliveryTime) { String zipCode = deliveryAddress.getZip(); int dayOfWeek = DateTimeUtil.dayOfWeek(deliveryTime); int timeOfDay = DateTimeUtil.timeOfDay(deliveryTime); String closingTimesKey = closingTimesKey(zipCode, dayOfWeek);
Set<String> trsClosingAfter = redisTemplate.opsForZSet().rangeByScore(closingTimesKey, timeOfDay, 2359);
Set<String> restaurantIds = new HashSet<String>(); for (String tr : trsClosingAfter) { String[] values = tr.split("_"); if (Integer.parseInt(values[0]) <= timeOfDay) restaurantIds.add(values[1]); } Collection<String> keys = keyFormatter.keys(restaurantIds); return availableRestaurantTemplate.opsForValue().multiGet(keys); }
Find those that close after
Filter out those that open after
Retrieve open restaurants
@crichardson
Sorry Ted!
http://en.wikipedia.org/wiki/Edgar_F._Codd
@crichardson
Agenda
• Why polyglot persistence?
• Using Redis as a cache
• Optimizing queries using Redis materialized views
• Synchronizing MySQL and Redis
• Tracking changes to entities
• Using a modular asynchronous architecture
@crichardson
MySQL & Redis need to be consistent
@crichardson
Two-Phase commit is not an option
• Redis does not support it
• Even if it did, 2PC is best avoided http://www.infoq.com/articles/ebay-scalability-best-practices
@crichardson
AtomicConsistentIsolatedDurable
Basically AvailableSoft stateEventually consistent
BASE: An Acid Alternative http://queue.acm.org/detail.cfm?id=1394128
@crichardson
Updating Redis #FAILbegin MySQL transaction update MySQL update Redisrollback MySQL transaction
Redis has updateMySQL does not
begin MySQL transaction update MySQLcommit MySQL transaction<<system crashes>> update Redis
MySQL has updateRedis does not
@crichardson
Updating Redis reliablyStep 1 of 2
begin MySQL transactionupdate MySQLqueue CRUD event in MySQL
commit transaction
ACID
Event IdOperation: Create, Update, DeleteNew entity state, e.g. JSON
@crichardson
Updating Redis reliably Step 2 of 2
for each CRUD event in MySQL queue get next CRUD event from MySQL queue
If CRUD event is not duplicate then Update Redis (incl. eventId)end ifbegin MySQL transaction
mark CRUD event as processedcommit transaction
@crichardson
ID JSON processed?
ENTITY_CRUD_EVENT
EntityCrudEventProcessor
RedisUpdater
apply(event)
Step 1 Step 2
EntityCrudEventRepository
INSERT INTO ...
Timer
SELECT ... FROM ...
Redis
@crichardson
Updating Redis
WATCH restaurant:lastSeenEventId:≪restaurantId≫
Optimistic locking
Duplicate detection
lastSeenEventId = GET restaurant:lastSeenEventId:≪restaurantId≫
if (lastSeenEventId >= eventId) return;
Transaction
MULTI SET restaurant:lastSeenEventId:≪restaurantId≫ eventId
... update the restaurant data...
EXEC
@crichardson
Agenda
• Why polyglot persistence?
• Using Redis as a cache
• Optimizing queries using Redis materialized views
• Synchronizing MySQL and Redis
• Tracking changes to entities
• Using a modular asynchronous architecture
@crichardson
How do we generate CRUD events?
@crichardson
Change tracking options
• Explicit code
• Hibernate event listener
• Service-layer aspect
• CQRS/Event-sourcing
@crichardson
ID JSON processed?
ENTITY_CRUD_EVENT
EntityCrudEventRepository
HibernateEventListener
@crichardson
Hibernate event listenerpublic class ChangeTrackingListener implements PostInsertEventListener, PostDeleteEventListener, PostUpdateEventListener { @Autowired private EntityCrudEventRepository entityCrudEventRepository; private void maybeTrackChange(Object entity, EntityCrudEventType eventType) { if (isTrackedEntity(entity)) { entityCrudEventRepository.add(new EntityCrudEvent(eventType, entity)); } }
@Override public void onPostInsert(PostInsertEvent event) { Object entity = event.getEntity(); maybeTrackChange(entity, EntityCrudEventType.CREATE); }
@Override public void onPostUpdate(PostUpdateEvent event) { Object entity = event.getEntity(); maybeTrackChange(entity, EntityCrudEventType.UPDATE); }
@Override public void onPostDelete(PostDeleteEvent event) { Object entity = event.getEntity(); maybeTrackChange(entity, EntityCrudEventType.DELETE); }
@crichardson
Agenda
• Why polyglot persistence?
• Using Redis as a cache
• Optimizing queries using Redis materialized views
• Synchronizing MySQL and Redis
• Tracking changes to entities
• Using a modular asynchronous architecture
@crichardson
Original architecture
WAR
RestaurantManagement
...
@crichardson
Drawbacks of this monolithic architecture
• Obstacle to frequent deployments
• Overloads IDE and web container
• Obstacle to scaling development
• Technology lock-in
WAR
RestaurantManagement
...
@crichardson
Need a more modular architecture
@crichardson
Using a message broker
Asynchronous is preferred
JSON is fashionable but binary format is more efficient
@crichardson
Modular architecture
Order taking
Restaurant Management
MySQL Database
RedisCacheRedis RabbitMQ
CONSUMER RESTAURANT OWNER
Event Publisher
Timer
@crichardson
Benefits of a modular asynchronous architecture
• Scales development: develop, deploy and scale each service independently
• Redeploy UI frequently/independently
• Improves fault isolation
• Eliminates long-term commitment to a single technology stack
• Message broker decouples producers and consumers
@crichardson
Step 2 of 2
for each CRUD event in MySQL queue get next CRUD event from MySQL queue
Publish persistent message to RabbitMQbegin MySQL transaction
mark CRUD event as processedcommit transaction
@crichardson
Message flowEntityCrudEvent
Processor
RedisUpdater
RABBITMQ
AvailableRestaurantManagementService
REDIS
Spring Integration glue code
@crichardson
RedisUpdater ⇒ AMQP<beans>
<int:gateway id="redisUpdaterGateway" service-interface="net...RedisUpdater" default-request-channel="eventChannel" />
<int:channel id="eventChannel"/>
<int:object-to-json-transformer input-channel="eventChannel" output-channel="amqpOut"/>
<int:channel id="amqpOut"/>
<amqp:outbound-channel-adapter channel="amqpOut" amqp-template="rabbitTemplate" routing-key="crudEvents" exchange-name="crudEvents" />
</beans>
Creates proxy
@crichardson
AMQP ⇒ Available...Service<beans> <amqp:inbound-channel-adapter channel="inboundJsonEventsChannel" connection-factory="rabbitConnectionFactory" queue-names="crudEvents"/>
<int:channel id="inboundJsonEventsChannel"/> <int:json-to-object-transformer input-channel="inboundJsonEventsChannel" type="net.chrisrichardson.foodToGo.common.JsonEntityCrudEvent" output-channel="inboundEventsChannel"/> <int:channel id="inboundEventsChannel"/> <int:service-activator input-channel="inboundEventsChannel" ref="availableRestaurantManagementServiceImpl" method="processEvent"/></beans>
Invokes service
@crichardson
Summary
• Each SQL/NoSQL database = set of tradeoffs
• Polyglot persistence: leverage the strengths of SQL and NoSQL databases
• Use Redis as a distributed cache
• Store denormalized data in Redis for fast querying
• Reliable database synchronization required
@crichardson
@crichardson [email protected]://plainoldobjects.com - code and slides
Sign up for CloudFoundry.com