MongoDB for Time Series Data: Sharding
Download as PPTX, PDF4 likes3,639 views
1) The document discusses sharding time series sensor data from 16,000 traffic sensors across the US to support a nationwide traffic monitoring application. 2) It models the read, write and storage patterns and determines that a sharded cluster is needed to store over 500GB of yearly data that will grow significantly over time. 3) It recommends using a compound shard key of {linkID, date} to distribute writes evenly while enabling targeted queries, and storing summary data in a separate replica set for performance.
![{ _id: “900006:2014031206”,
data: [
{ speed: NaN, time: NaN },
{ speed: NaN, time: NaN },
{ speed: NaN, time: NaN },
...
],
conditions: {
status: "unknown",
pavement: "unknown",
weather: "unknown"
}
}
Sample Document Structure
Pre-allocated,
60 element array of
per-minute data](https://image.slidesharecdn.com/jakesshardingtimeseriesdatawebinar20140715-140716141539-phpapp01/85/MongoDB-for-Time-Series-Data-Sharding-9-320.jpg)
![Replica Set Tags
• app servers in different data centers use
replica set tags plus read preferencenearest
• db.collection.find().readPref( { mode: 'nearest',
tags: [ {'datacenter': 'east'} ] } )
east
Secondary
Secondary
Primary
>rs.conf()
{"_id":"rs0",
"version":2,
"members":[
{"_id":0,
"host":"node0.example.net:27017",
"tags":{"datacenter":"east"}
},
{"_id":1,
"host":"node1.example.net:27017",
"tags":{"datacenter":"east"}
},
{"_id":2,
"host":"node2.example.net:27017",
"tags":{"datacenter":"east"}
},
}](https://image.slidesharecdn.com/jakesshardingtimeseriesdatawebinar20140715-140716141539-phpapp01/85/MongoDB-for-Time-Series-Data-Sharding-19-320.jpg)
MongoDB for Time Series Data: Sharding
- 1. Sr. Solutions Architect, MongoDB Jake Angerman Sharding Time Series Data
- 2. Let's Pretend We Are DevOps What my friends think I do What society thinks I do What my Mom thinks I do What my boss thinks I do What I think I do What I really do DevOps
- 3. Sharding Overview Primary Secondary Secondary Shard 1 Primary Secondary Secondary Shard 2 Primary Secondary Secondary Shard 3 Primary Secondary Secondary Shard N … Query Router Query Router Query Router …… Driver Application
- 4. Why do we need to shard? • Reaching a limit on some resource – RAM (working set) – Disk space – Disk IO – Client network latency on writes (tag aware sharding) – CPU
- 5. Do we need to shard right now? • Two schools of thought: 1. Shard at the outset to avoid technical debt later 2. Shard later to avoid complexity and overhead today • Either way, shard before you need to! – 256GB data size threshold published in documentation – Chunk migrations can cause memory contention and disk IO Working Set Free RAM Things seemed fine… Working Set Chunk Migration … then I waited too long to shard
- 6. Develop Nationwide traffic monitoring system
- 7. Traffic sensors to monitor interstate conditions • 16,000 sensors • Measure • Speed • Travel time • Weather, pavement, and traffic conditions • Support desktop, mobile, and car navigation systems
- 8. Model After NY State Solution http://511ny.org
- 9. { _id: “900006:2014031206”, data: [ { speed: NaN, time: NaN }, { speed: NaN, time: NaN }, { speed: NaN, time: NaN }, ... ], conditions: { status: "unknown", pavement: "unknown", weather: "unknown" } } Sample Document Structure Pre-allocated, 60 element array of per-minute data
- 10. > db.mdbw.stats() { "ns" : "test.mdbw", "count" : 16000, // one hour's worth of documents "size" : 65280000, // size of user data, padding included "avgObjSize" : 4080, "storageSize" : 93356032, // size of data extents, unused space included "numExtents" : 11, "nindexes" : 1, "lastExtentSize" : 31354880, "paddingFactor" : 1, "systemFlags" : 1, "userFlags" : 1, "totalIndexSize" : 801248, "indexSizes" : { "_id_" : 801248 }, "ok" : 1 } collection stats
- 11. Storage model spreadsheet sensors 16,000 years to keep data 6 docs per day 384,000 docs per year 140,160,000 docs total across all years 840,960,000 indexes per day 801248 bytes storage per hour 63 MB storage per day 1.5 GB storage per year 539 GB storage across all years 3,235 GB
- 12. Why we need to shard now 539 GB in year one alone 0 500 1,000 1,500 2,000 2,500 3,000 3,500 1 2 3 4 5 6 Year Total storage (GB) 16,000 sensors today… … 47,000 tomorrow?
- 13. What will our sharded cluster look like? • We need to model the application to answer this question • Model should include: – application write patterns (sensors) – application read patterns (clients) – analytic read patterns – data storage requirements • Two main collections – summary data (fast query times) – historical data (analysis of environmental conditions)
- 14. Option 1: Everything in one sharded cluster Primary Primary Primary Secondary Secondary Secondary Secondary Secondary Secondary Shard 2 Shard 3 Shard N … Primary Secondary Secondary Shard 1 Primary Shard Primary Secondary Secondary Shard 4 • Issue: prevent analytics jobs from affecting application performance • Summary data is small (16,000 * N bytes) and accessed frequently
- 15. Option 2: Distinct replica set for summaries Primary Primary Primary Secondary Secondary Secondary Secondary Secondary Secondary Shard 1 Shard 2 Shard N … Primary Secondary Secondary Replica set Primary Secondary Secondary Shard 3 • Pros: Operational separation between business functions • Cons: application must write to two different databases
- 16. Application read patterns • Web browsers, mobile phones, and in-car navigation devices • Working set will be kept in RAM • 5M subscribers * 1% active * 50 sensors/query * 1 device query/min = 41,667 reads/sec • 41,667 reads/sec * 4080 bytes = 162 MB/sec – and that's without any protocol overhead • Gigabit Ethernet is ≈ 118 MB/sec Primary Secondary Secondary Replica set 1 Gbps
- 17. Application read patterns (continued) • Options – provision more bandwidth ($$$) – tune application read pattern – add a caching layer – secondary reads from the replica set Primary Secondary Secondary Replica set 1 Gbps 1 Gbps 1 Gbps
- 18. Secondary Reads from the Replica Set • Stale data OK in this use case • caution: read preference of secondary could be disastrous in a 3-replica set if a secondary fails! • app servers with mixed read preferences of primary and secondary are operationally cumbersome • Use nearest read preference to access all nodes Primary Secondary Secondary Replica set 1 Gbps 1 Gbps 1 Gbps db.collection.find().readPref ( { mode: 'nearest'} )
- 19. Replica Set Tags • app servers in different data centers use replica set tags plus read preferencenearest • db.collection.find().readPref( { mode: 'nearest', tags: [ {'datacenter': 'east'} ] } ) east Secondary Secondary Primary >rs.conf() {"_id":"rs0", "version":2, "members":[ {"_id":0, "host":"node0.example.net:27017", "tags":{"datacenter":"east"} }, {"_id":1, "host":"node1.example.net:27017", "tags":{"datacenter":"east"} }, {"_id":2, "host":"node2.example.net:27017", "tags":{"datacenter":"east"} }, }
- 20. eastcentralwest Replica Set Tags • Enables geographic distribution SecondarySecondary Primary
- 21. eastcentralwest Replica Set Tags • Enables geographic distribution • Allows scaling within each data center Secondary Secondary Secondary Secondary Secondary Secondary Primary Secondary Secondary
- 22. Analytic read patterns • How does an analyst look at the data on the sharded cluster? • 1 Year of data = 539 GB 3, 256 3, 192 5, 128 9, 64 17, 32 0 50 100 150 200 250 300 0 2 4 6 8 10 12 14 16 18 Server RAM Number of shards
- 23. Application write patterns • 16,000 sensors every minute = 267 writes/sec • Could we handle 16,000 writes in one second? – 16,000 writes * 4080 bytes = 62 MB • Load test the app!
- 24. Modeling the Application - summary • We modeled: – application write patterns (sensors) – application read patterns (clients) – analytic read patterns – data storage requirements – the network, a little bit
- 25. Shard Key
- 26. Shard Key characteristics • Agood shard key has: – sufficient cardinality – distributed writes – targeted reads ("query isolation") • Shard key should be in every query if possible – scatter gather otherwise • Choosing a good shard key is important! – affects performance and scalability – changing it later is expensive
- 27. Hashed shard key • Pros: – Evenly distributed writes • Cons: – Random data (and index) updates can be IO intensive – Range-based queries turn into scatter gather Shard 1 mongos Shard 2 Shard 3 Shard N
- 28. Low cardinality shard key • Induces "jumbo chunks" • Examples: sensor ID • Makes sense for some use cases besides this one Shard 1 mongos Shard 2 Shard 3 Shard N [ a, b ) [ b, c ) [ c, d ) [ e, f )
- 29. Ascending shard key • Monotonically increasing shard key values cause "hot spots" on inserts • Examples: timestamps, _id Shard 1 mongos Shard 2 Shard 3 Shard N [ ISODate(…), $maxKey )
- 30. Choosing a shard key for time series data • Consider compound shard key: {arbitrary value, incrementing value} • Best of both worlds – multi-hot spotting, targeted reads Shard 1 mongos Shard 2 Shard 3 Shard N [ {V1, ISODate(A)}, {V1, ISODate(B)} ), [ {V1, ISODate(B)}, {V1, ISODate(C)} ), [ {V1, ISODate(C)}, {V1, ISODate(D)} ), … [ {V4, ISODate(A)}, {V4, ISODate(B)} [ {V4, ISODate(B)}, {V4, ISODate(C)} [ {V4, ISODate(C)}, {V4, ISODate(D)} … [ {V2, ISODate(A)}, {V2, ISODate(B)} ), [ {V2, ISODate(B)}, {V2, ISODate(C)} ), [ {V2, ISODate(C)}, {V2, ISODate(D)} ), … [ {V3, ISODate(A)}, {V3, ISODate(B)} ), [ {V3, ISODate(B)}, {V3, ISODate(C)} ), [ {V3, ISODate(C)}, {V3, ISODate(D)} ), …
- 31. What is our shard key? • Let's choose: linkID, date – example: { linkID: 9000006, date: 2014031206 } – example: { _id: "900006:2014031206" } – this application's _id is in this form already, yay!
- 32. Summary • Model the read/write patterns and storage • Choose an appropriate shard key • DevOps influenced the application – write recent summary data to separate database – replica set tags for summary database – avoid synchronous sensor checkins – consider changing client polling frequency – consider throttling RESTAPI access to app servers
- 33. Sign up for our “Path to Proof” Program and get free expert advice on implementation, architecture, and configuration. www.mongodb.com/lp/contact/path-proof-program
- 35. Sr. Solutions Architect, MongoDB Jake Angerman Thank You
Editor's Notes
- #6: Public Service Announcement need to research the 256 GB limit
- #10: Compound unique index on linkId & Interval update field used to identify new documents for aggregation
- #21: you could use the same config file in each data center! 47K is 3X 16K
- #22: application must read a config file to determine its read pool 47K is 3X 16K
- #29: may consider hashing _id instead
- #30: may consider hashing _id instead