Slide 1

Slide 1 text

MULTI-DIMENSIONAL SCALING – A NEW ARCHITECTURE FOR SCALING BIG DATA APPLICATION NoSQL Night Singapore

Slide 2

Slide 2 text

©2015  Couchbase  Inc.   2   Agenda   § Brief  history  of  scaling  in  database   §  Scaling  up     §  Scaling  out   § NoSQL  workloads  and  scalability  model   §  Core  data  operations,  indexing  and  querying   §  Homogenous  HW  scaling   § Introducing  multi-­‐dimensional  scaling   §  Isolation  of  workloads  through  a  services  architecture   §  Independent  HW  scaling   § Q  &  A  

Slide 3

Slide 3 text

©2015  Couchbase  Inc.   3   Disclaimer  &  Notes   §  Disclaimer:  The  views  expressed  in  this  presentation  are  our  own  and  do   not  necessarily  reflect  the  views  of  Couchbase   §  Most  of  the  content  in  this  presentation  is  originally  created  by  Anil,   Product  Manager,  Couchbase   §  Thanks  Anil,  for  allowing  me  to  re-­‐use  your  slides  J     3  

Slide 4

Slide 4 text

©2015  Couchbase  Inc.   4   About  the  Speaker   § Clarence J M Tauro – [email protected] -  Senior Instructor, Couchbase -  ~11 Years Professional Teaching and Consulting Experience -  Worked at Pivotal – Instructor/Consultant for Spring/Spring Security/Spring Web/Enterprise Integration with Spring/Spring JMS/Spring Web/Spring Batch, Pivotal Hadoop/Cloud Foundry -  PhD in Computer Science from Christ University [thesis accepted] -  Hard-core Dog lover 4  

Slide 5

Slide 5 text

Brief  History  of  Scaling  in  Database  

Slide 6

Slide 6 text

©2015  Couchbase  Inc.   6   Scaling  up   § Scale-­‐up  architecture   §  Cluster  processors  –  hyper-­‐threading  to  cores   §  Locally  partition  workload  among  processors     §  Communicate  over  memory  

Slide 7

Slide 7 text

©2015  Couchbase  Inc.   7   Scaling  Up  –  Pros  &  Cons   Pros   §  Can  result  in  major   performance  improvement   §  Machines  can  now  support   having  many  cores  and   terabytes  of  RAM     Cons   §  Expensive   §  Requires  downtime   §  Performance  bounded  –  at   some  point  the  database   engine  itself  becomes  the   bottleneck     §  Limited  in  scalability  and   elasticity    

Slide 8

Slide 8 text

©2015  Couchbase  Inc.   8   Scaling  out   § Scale-­‐out  architecture   §  Cluster  of  commodity  HW     §  Horizontal  partitioning  of  data  on  cluster  nodes   §  Communicate  over  network  

Slide 9

Slide 9 text

©2015  Couchbase  Inc.   9   Scaling  Out  –  Pros  &  Cons   Pros   §  Simple  and  easily  scalable   §  Data  evenly  split  across  cluster   of  nodes   §  Scales  linearly  with  throughput   §  Highly  available   §  No  single  point  of  failure   Cons   §  Not  great  for  all  workloads  – data,  index  and  query    

Slide 10

Slide 10 text

©2015  Couchbase  Inc.   10   So  Which  Model  is  the  Right  Model  ?   Scale  up  or  scale  out?  

Slide 11

Slide 11 text

©2015  Couchbase  Inc.   11   Scaling  Up  vs.  Scale  Out   Link  to  whitepaper  -­‐   http://www.msr-­‐ waypoint.com/pubs/204499/ a20-­‐appuswamy.pdf      

Slide 12

Slide 12 text

NoSQL  Workloads  &  Scalability  Model  

Slide 13

Slide 13 text

©2015  Couchbase  Inc.   13   NoSQL  Workloads   § One  database,  many  workloads     §  Core  data  processing:  GETs  &  SETs  for  a  given  key   §  Indexing:  Index  maintenance  and  lookups   §  Querying:  Combine  index  and  data  with  complex  just-­‐in-­‐time  data   re-­‐shaping,  ordering,  grouping,  aggregations,  and  more     Varying  resource  requirements  -­‐  CPU,  RAM,  I/O,  Network   Varying  methods  to  optimize  latency  &  throughput  for  each  

Slide 14

Slide 14 text

©2015  Couchbase  Inc.   14   Scalability  Model  Today   Homogenous  Scaling   §  Each  node  get  a  slice  of  the  workload   §  Simple  to  do…   But...   §  Workloads  compete  and  interfere  with  each  other   §  Can’t  fine  tune  each  workload   §  Core  Data  operation  are  partition-­‐able  so  great  with  wider  fan-­‐out   §  Indexing  and  queries  aren’t  always  partitionable,  so  worse  with  wider  fan-­‐out     Index  Workload   Couchbase  Cluster   Query  Workload   Data  Workload   node1   node8  

Slide 15

Slide 15 text

Introducing  Multi-­‐Dimensional  Scaling  

Slide 16

Slide 16 text

©2015  Couchbase  Inc.   16   Modern  Architecture   What  is  Multi-­‐Dimensional  Scalability?     MDS  is  the  architecture  that  enables  independent  scaling  of   data,  query  and  indexing  workloads.       Index  Service   Couchbase  Cluster   Query  Service   Data  Service   node1   node8  

Slide 17

Slide 17 text

©2015  Couchbase  Inc.   17   Couchbase  Cluster   node1   node8   Modern  Architecture   § Isolated  Service  for  minimized  interference   §  Independent  “zones”  for  Query,  Index  and  Data  Services   Minimize  indexing  and  query  overhead  on  core  key-­‐value  operations.   Index  Service   Query  Service   Data  Service  

Slide 18

Slide 18 text

©2015  Couchbase  Inc.   18   Modern  Architecture   §  Independent  Scalability  for  Best  Computational  Capacity  per  Service       Heavier  indexing  (index  more  fields)  :  scale  up  index  service  nodes    More  RAM  for  query  processing:  scale  up  query  service  nodes       Couchbase  Cluster   node1   node8   node9   Data  Service   Index  Service   Query  Service  

Slide 19

Slide 19 text

Services  Architecture   Data,  Index,  &  Query  

Slide 20

Slide 20 text

©2015  Couchbase  Inc.   20   Full  Cluster  Architecture   STORAGE   Couchbase  Server  1   SHARD   7   SHARD   9   SHARD   5   SHARD   SHARD     SHARD     Managed   Cache   Cluster     Manager   Cluster   Manager   Managed     Cache   Storage     Data  Service   Index  Service   Query  Service   STORAGE   Couchbase  Server  2   SHARD   7   SHARD   9   SHARD   5   SHARD   SHARD     SHARD     Managed   Cache   Cluster     Manager   Cluster   Manager   Managed     Cache   Storage     Data  Service   Index  Service   Query  Service   STORAGE   Couchbase  Server  3   SHARD   7   SHARD   9   SHARD   5   SHARD   SHARD     SHARD     Managed   Cache   Cluster     Manager   Cluster   Manager   Managed     Cache   Storage     Data  Service   Index  Service   Query  Service   STORAGE   Couchbase  Server  4   SHARD   7   SHARD   9   SHARD   5   SHARD   SHARD     SHARD     Managed   Cache   Cluster     Manager   Cluster   Manager   Managed     Cache   Storage     Data  Service   Index  Service   Query  Service   STORAGE   Couchbase  Server  5   SHARD   7   SHARD   9   SHARD   5   SHARD   SHARD     SHARD     Managed   Cache   Cluster     Manager   Cluster   Manager   Managed     Cache   Storage     Data  Service   Index  Service   Query  Service   STORAGE   Couchbase  Server  6   SHARD   7   SHARD   9   SHARD   5   SHARD   SHARD     SHARD     Managed   Cache   Cluster     Manager   Cluster   Manager   Managed     Cache   Storage     Data  Service   Index  Service   Query  Service  

Slide 21

Slide 21 text

©2015  Couchbase  Inc.   21   Full  Cluster  Architecture   STORAGE   Couchbase  Server  1   SHARD   7   SHARD   9   SHARD   5   SHARD   SHARD     SHARD     Managed   Cache   Cluster     Manager   Cluster   Manager   Managed     Cache   Storage     Data  Service   Index  Service   Query  Service   STORAGE   Couchbase  Server  2   SHARD   7   SHARD   9   SHARD   5   SHARD   SHARD     SHARD     Managed   Cache   Cluster     Manager   Cluster   Manager   Managed     Cache   Storage     Data  Service   Index  Service   Query  Service   STORAGE   Couchbase  Server  3   SHARD   7   SHARD   9   SHARD   5   SHARD   SHARD     SHARD     Managed   Cache   Cluster     Manager   Cluster   Manager   Managed     Cache   Storage     Data  Service   Index  Service   Query  Service   STORAGE   Couchbase  Server  4   SHARD   7   SHARD   9   SHARD   5   SHARD   SHARD     SHARD     Managed   Cache   Cluster     Manager   Cluster   Manager   Managed     Cache   Storage     Data  Service   Index  Service   Query  Service   STORAGE   Couchbase  Server  5   SHARD   7   SHARD   9   SHARD   5   SHARD   SHARD     SHARD     Managed   Cache   Cluster     Manager   Cluster   Manager   Managed     Cache   Storage     Data  Service   Index  Service   Query  Service   STORAGE   Couchbase  Server  6   SHARD   7   SHARD   9   SHARD   5   SHARD   SHARD     SHARD     Managed   Cache   Cluster     Manager   Cluster   Manager   Managed     Cache   Storage     Data  Service   Index  Service   Query  Service  

Slide 22

Slide 22 text

©2015  Couchbase  Inc.   22   Storage  Engine   Cluster  Manager   Data  Service                 Projector  &  Router   New  Services  in  Couchbase  Server  4.0   Query  Service                 Index  Service                 Supervisor   Index  maintenance  &     Scan  coordinator   Index#2   Index#1   Query  Processor   cbq-­‐engine   Bucket#1   Bucket#2   DCP  Stream   Index#4   Index#3   ...   Bucket#2   18093   11211   18901   Managed  Cache  

Slide 23

Slide 23 text

©2015  Couchbase  Inc.   23   Recap   § MDS  enables  unprecedented  control  of  scalability  with   Couchbase  Server   §  Separate  out  competing  workloads  to  independent  services   §  Independently  scale  each  service  “zone”  within  the  cluster   § Couchbase  Server  with  MDS  maximizes  scalability  and   performance   §  Improves  scale  and  performance  to  degrees  not  possible  with  other   NoSQL  or  big-­‐data  engines  on  premise  or  in  the  cloud   §  Improved  price/performance  and  squeezes  more  performance  and   throughput  for  mission-­‐critical  systems  

Slide 24

Slide 24 text

Thank  you.  

Slide 25

Slide 25 text

Get  Started  with  Couchbase  Server  4.0:  www.couchbase.com/beta   Get  Trained  on  Couchbase:  http://training.couchbase.com       CD220: Developing Couchbase NoSQL Applications Oct 20 – Oct 23 2015 CS300: Couchbase NoSQL Server Administration Nov 17 – Nov 20 Enroll Today!