![HSET dog:1 name Honey breed Greyhound
Command Name
Key Name
Field Name 1
Field Value 1
Field Name 2
Field Value 2
db.HashSet("dog:1", new HashEntry[] {new ("name", "Honey"), new
("breed", "Greyhound")});](https://image.slidesharecdn.com/anintroductiontoredisfor-220722133724-3f3be438/85/An-Introduction-to-Redis-for-NET-Developers-pdf-63-320.jpg)
An Introduction to Redis for .NET Developers.pdf
- 1. An Introduction to Redis for .NET Developers Presented By: Steve Lorello @slorello
- 3. Agenda ● What is Redis? ● Using Redis in your app ● Major Redis Use Cases ● Redis Data Structures ● Redis Design Patterns ● Redis Gotchas and Anti-Patterns
- 5. ● 80% ecosystem neutral ● 20% .NET focused ● StackExchange.Redis - Canonical Client
- 6. What is ?
- 7. What is Redis? ● NOSQL Database ● REmote DIctionary Server
- 8. What is Redis? ● Completely In Memory ● Key-Value Data Structure Store
- 9. What is Redis? ● Redis is Single Threaded ● Blazingly Fast
- 10. ● Easy to Use ● Beloved by Developers What is Redis?
- 11. Redis is ACIDic
- 12. ACID in Redis - Atomicity ● Redis is Single Threaded ● Redis commands are completely atomic ● ‘Transactions’ & Scripting for grouping commands
- 13. ACID in Redis - Consistency ● Depends on deployment and config ● Single instance always consistent ● Read-only replication guaranteed eventual consistency ○ Forced consistency with “WAIT” command
- 14. ACID in Redis - Isolation ● Single Threaded ● Isolated as there’s no concurrency
- 15. ACID in Redis - Durability ● Entire database loaded into memory ● Two durability models persistence to disk ● AOF (Append Only File) ● RDB (Redis Database File)
- 16. Durability - Append Only File (AOF) ● With each command Redis writes to AOF ● Reconstruct Redis from AOF ● AOF flush to disk is NOT necessarily synchronous ● FSYNC config policy determines level of durability ○ always - Synchronously flush to disk (fully durable) ○ everysec - Flush buffer to disk every second ○ no - OS decides when to write to disk
- 17. Durability - Redis Database file (RDB) ● RDBs are snapshots of the database ● Taken in intervals, or by command ● More compact and faster to ingest than AOF ● Less strain on OS than AOF ● Comes at cost of higher potential data loss
- 19. ConnectionMultiplexer ● Heart of StackExchange.Redis ● Handles connections to Redis ● Multiplexes all Commands Sent to Redis ● Initialize 1 instance/app ○ Pass around via DI or can be used as a Singleton
- 20. IDatabase ● Transient handle to Redis Commands ● ConnectionMultiplexer.GetDatabase() ● Intalize/destroy at will
- 21. IBatch ● Group of commands ○ Sent to Redis as one continuous unit ○ NOT guaranteed to be executed by Redis sequentially ● IDatbase.CreateBatch
- 23. Redis Standalone
- 24. Redis Standalone ● One Redis instance ● Simple ● No HA capabilities ● No Horizontal Scaling
- 25. Connecting ● Basic Connection String ● {host}:{port},password=password,ssl=sslSetting redis:6379,user=default,password=password,ssl=false
- 27. Redis Sentinel
- 28. Sentinel, Replication for High Availability ● Writable Master ● Read-only Replicas ● “Sentinels” to handle HA & Failover
- 29. Sentinel (Replication for High Availability) Source: https://www.tecmint.com/setup-redis-high-availability-with-sentinel-in-centos-8/
- 30. Connection ● Connect to sentinel, not to a master ● Pass in serviceName ○ Signals that it’s a sentinel sentinel:26379,servcieName=mySentinel,password=password
- 31. Redis Cluster
- 32. Redis Cluster - Horizontal Scaling - ‘Hard Mode’ ● Keyspace split across multiple shards ● Deterministic hash function on key to determine ‘slot’ ● Shards responsible for group of ‘slots’
- 33. Redis Cluster ● Horizontally scale reads/writes across cluster ● Multi-key operations become slot limited
- 34. Connecting ● Similar to Standalone ● You can pass in multiple endpoints ○ Helps in case of failovers redis-1:6379,redis-2:6379,redis-3:6379,ssl=true,password=password
- 36. Redis Cloud
- 37. Redis Cloud - Infinite Scaling ‘Easy Mode’ ● Have Redis manage your cluster ● Scales endlessly ● Zero out complexity of scaling Redis
- 38. Other Redis Cloud Features (pay the bills slide) ● Geo-Distribution with Active-Active ● Tremendous throughput ● Five-nines availability ● Multi-cloud product ● Access to Redis Modules ● Enterprise support
- 39. Redis Cloud - A Developer Advocate’s Experience ● We run a discord server Join us by the way! https://discord.gg/redis ● Field lots of complex questions about cluster/sentinel from the community ● Write up lots of complex answers with the typical refrain: ○ “Or you could just use Redis Cloud and you don’t have to worry about any of this”
- 40. Uses of Redis
- 41. As a Cache ● #1 Most Popular Usage ● Low Latency Data Access ● For repeated operations
- 42. As a Session State Store ● Distributed Session State ● Low-Latency Access
- 43. As a Message Bus ● Stream Messages Through Redis with Redis Streams ● Consume message with consumer groups ● “Using Redis Streams saved our company from Bankruptcy”
- 44. As a Database ● No extra layers of caching, just use cache as DB! ● Document Data Structures allow for CRUD operations ● Indexing Capabilities to easily find things
- 46. Strings
- 47. Strings ● Simplest Redis Data Structure ● Single Key -> Single Value ● Driven primarily by SET/GET like commands
- 48. SET foo bar EX 50 Command Name Key Name Value Expire Option Expire in 50 seconds db.StringSet("foo", "bar", TimeSpan.FromSeconds(50));
- 49. GET foo Command Name Key Name db.StringGet("foo");
- 51. Sets ● Mathematical set ● Collection of Redis Strings ● Unordered ● No Duplication ● Set Combination Operations (Intersect, Union, Difference)
- 52. SADD myset thing-1 Command Name Key Name Value db.SetAdd("myset1", "thing-1");
- 53. SISMEMBER myset thing-1 Command Name Key Name Value db.SetContains("myset", "thing-1");
- 54. SMEMBERS myset Command Name Key Name db.SetMembers("myset");
- 55. SUNION myset myotherset Command Name Key 1 Key 2 db.SetCombine(SetOperation.Union,"myset","myset2");
- 56. Sorted Sets
- 57. Sorted Sets ● Same membership rules as sets ● Set Sorted by Member ‘score’ ● Can be read in order or reverse order
- 58. Sorted Set Range Types ● By Rank ● By Score ● By Lex
- 59. ZADD users:last_visited 1651259876 Steve Command Name Key Name Score Value db.SortedSetAdd("users:last_visited", "steve", 1651259876 );
- 60. ZRANGE users:last_visited 0 -1 Key Name Command Name Start Value db.SortedSetRangeByRank("users:last_visited");
- 61. Hashes
- 62. Hashes ● Field value stores ● Appropriate for flat objects ● Field Names and Values are Redis Strings
- 63. HSET dog:1 name Honey breed Greyhound Command Name Key Name Field Name 1 Field Value 1 Field Name 2 Field Value 2 db.HashSet("dog:1", new HashEntry[] {new ("name", "Honey"), new ("breed", "Greyhound")});
- 64. HGET dog:1 name Command Name Key Name Field Name db.HashGet("dog:1", "name");
- 65. JSON
- 66. JSON ● Store full JSON objects in Redis ● Access fields of object via JSON Path ● Redis Module
- 67. JSON.SET dog:1 $ ‘{“name” : “Honey”, “breed” : “Greyhound”}’ Command Name Key Name Path JSON db.Execute("JSON.SET", "dog:1", "$", "{"name":"Honey","breed":"Greyhound"}");
- 68. JSON.GET dog:1 $.breed Command Name Key Name Path db.Execute("JSON.GET", "dog:1", "$.breed");
- 69. Streams
- 70. Streams ● Implements Log Data Structure ● Send messages through Streams as Message Queue ● Messages Consist of ID & list of field-value pairs ● Consumer Groups for coordinated reads
- 72. Special IDs Id Command Description * XADD Auto-Generate ID $ XREAD Only retrieve new messages > XREADGROUP Next Message for Group - XRANGE Lowest + XRANGE Highest
- 73. XADD sensor:1 * temp 27 Command Name Key Name Id Field Name 1 Field Value 1 db.StreamAdd("sensor:1", "temp", 27);
- 74. XREAD STREAMS sensor:1 $ Command Name Key Name Id db.StreamRead("sensor:1", 0);
- 75. XREADGROUP GROUP avg con1 sensor:1 > Command Name Key Name Id Group Name Consumer Name db.StreamReadGroup("sensor:1", "avg", "con1", ">");
- 78. The Root of Most Redis Bottlenecks ● Op-times in Redis Measured in Microseconds ● RTTs in Redis Measured in Milliseconds ● Minimize RTT ● Minimize number of Round Trips
- 79. Variadic Commands ● Many commands are variadic ● Leverage variadicity of commands
- 80. Always Be Pipelining ● Send multiple commands simultaneously ● Appropriate when intermediate results are unimportant ● Can dramatically increase throughput
- 81. Object Storage
- 82. Document Storage ● 3 Models of Document Storage ○ Hashes ○ Raw JSON ○ JSON Data Structure
- 83. Hashes ● Native flat-object Data structure ● HSET (which is variadic) to create/update ● HGET/HMGET/HGETALL to get fields in the hash ● HDEL/UNLINK to delete fields/objects
- 84. Pros: ● Native ● Performant Cons: ● Breaks down on more complicated objects ● Breaks down with collection storage
- 85. JSON Blobs ● Native data structure (a string) ● SET to create/update ● GET to read ● UNLINK to delete
- 86. Pros: ● Native ● Simple ● Your input is probably in this format Cons: ● Updates are expensive—O(N) ● Reads are expensive—O(N)
- 87. JSON Data Structure ● Exists in the Redis JSON module ● JSON.SET to create/update ● JSON.GET to read ● JSON.DEL to remove fields ● UNLINK to delete
- 88. Pros: ● All operations are fast ● Organized retrieval/update of data within object ● Works great with rich objects Cons: ● Needs a module
- 89. Finding Stuff
- 90. Indexing ● Finding by value requires the use of a pattern ● Two patterns ○ Indexing With Sorted Sets ○ Indexing With RediSearch
- 91. Sored Sets ● Construct Sorted sets to Index ● e.g. ○ Person:firstName:Steve{(0, Person:1)} ○ Person:age{(33, Person:1)} ● Query with ZRANGE commands ● Complex queries with Set Combination Commands
- 92. Pros: ● Native, in any Redis ● Performant Cons: ● Devilishly tricky to maintain consistency ● Cross-shard updates NOT atomic
- 93. RediSearch + Reids OM ● Predefine the Index in model ● Save insert objects into Redis: ○ collection.Insert(steve); ● Query with LINQ ○ collection.Where(x=>x.Name == “Steve”);
- 94. Pros: ● Fast ● Easy to use ● Cross-shard indexing taken care of Cons: ● Need to use a separate module
- 96. Transactions ● Apply multiple commands to server atomically ● Start with MULTI ● Issue commands ● End with EXEC ● WATCH keys to make sure they aren’t messed with during transaction ● If commands are valid Transaction will Execute
- 97. Lua Scripts ● Allow scripted atomic execution ● Allows you to use intermediate results atomically
- 98. local id = redis.call('incr', 'next_id') redis.call('set','key:' .. id, 'val') return id
- 99. Prepared Scripts ● StackExchange.Redis feature ● Tracks script sha ● Use @ to parameterize scripts ● Pass in object with parameters
- 100. var script = LuaScript.Prepare("return @name"); var scriptResult = db.ScriptEvaluate(script, new {name="Steve"});
- 101. Redis Gotchas and Anti-Patterns
- 102. ● #1 issue encountered ● Configurable timeout (defaults to 5 seconds) ● Clock starts when the library queues command internally ● Variety of causes ○ Slow Redis commands block redis ○ Very large payloads block connection ○ Thread pool starvation ○ Thread theft Timeouts
- 103. Reuse connections ● Connecting to Redis is Expensive ● Pool connection when possible ● Use one ConnectionMultiplexer
- 104. SCAN, please don’t use Keys ● KEYS command can lock up Redis ● Use cursor-based SCAN instead
- 105. Hot Keys ● Single keys with very heavy access patterns ● redis-cli --hotkeys ● Mitigates scaling advantages of clustered Redis ● Consider duplicating data across multiple keys ● Consider client-side caching
- 107. Resources Redis https://redis.io Redis-Stack docker Image https://hub.docker.com/r/redis/redis-stack/ StackExchange.Redis https://github.com/stackExchange/StackExchange.Redis Redis OM https://github.com/redis/redis-om-dotnet Code Examples https://github.com/slorello89/dotnet-redis-examples