buiding distributed system with akka #ScalaMatsuri
- anildigital
- anil wadghule
intro
- from india to japan
- anime
- sushi ramen
current scene in internet
- akka
size of internet
- growing
concurrent connection
- IoT
big data
cheap hardware
- cluster
- multiple
- fast network
bigdata streams of data
HA S fault
- High Scalable
- Fault Tlerant
- FB
- Reactive Principle
- Responsive
- Elaastic
- Resilient
- Message Driven
Distibuted System
- computer
- syncronized
- no global clock
- Leslie Lamport A guide to builindg
Tanenbaum and Steen Distibuted
Distributed Architecture
- Fallacies of Distributed System
- 0 Latency
- Bnadwidth
- network
- No Change Topology
Example
- Amazon
- Cassandra
- (no local splite db)
How do we scale
- Cofeeshop
- Regular small app
- replication/sahrding/consistent hasing
- replica head barista/assistant barista
- issue
- complexity
- consistency
- sharding
- ad more writers
- split orders with some key
- A-C J-L by key
- good tecnique for scaling
- issue
- limit data
- more complexity
- lmited data access patterns
- only good for certain kind of applications
- SAAS
- Consistent hassing
- consistent hasing and random trees
- e.g. akira to 9F72
- hash number
- any other
- node clash→replication factor hash number in some nodes
- consistency formula
- R + W > N
- total number of replica
- number of replicas acknowledge my update
- number of replicas
- scale transaction data
- CAP Theorem
- partition tolerance
- CAP partitiong not negotiated
- availability or consistency
distributed transactions
- ACID
- principles for database
- receive order process payment
paralellization/uneven workloads
scatter gatther
fault tolerance
- embraces the notion of failure
- components
- accounting for evey possible fault
- catch-all mecanism
- terminate as soon as uncaught failure
Javaconcurrent logs processor
- difficult to recover
- what we need is faulty component replaced in a threadable
- fault isolation
- structure
- redundancy
- replacement
- reboot
- suspend
resilience
- capacity to recover
- financial losses
actor model
- akka
- jvm actor model
- normal flow
- fault recovery flow
- filewatcher logprocessor dbwriter
- diskerror
- corrupt fileexception
- multiple
fualt isolation
- reboot
- structure
- redundancy
- replacement
- reboot
- component lifecycle
- suspend
akka
- akka cluster
- akka remote
- akka io
- akka core
- local actor
- remote actor
- scheduling
- routing
- cluster
- sharding
- persistence
- local actor →corrdinator actior
akka remoting
- actor on jvm 1 → actor on jvm 2
- port
akka cluster
- akka cluster
- fault tolerant membership service for akka
- no sngle point of failure or bottleneck
- akka cluster
- allows dynamically grow
- leader node node seed node
- akka cluster
- nodes
akka cluster features
- cluster membership
akka gossip protocol
- decentraised probabilistic
convergence
failure detector
- seed node
- up leaving unreachable exiting down
- removed
cluster singleton
cluster sharding distribute actors across several nodes
interact with them with local
shard corrdinator/shard region/shard
- akka dirstrbuted data
- akka cluster
- based on CRDTs
- less difficult
