Mohammad Umair Khan
Trainer, Mentor, Foo, Bar !
Architecture and Scalability
$ whoami
$ whoami
Software Developer
Jul 2013
Jul 2014
Snr. Software Developer
Jul 2015
Technical Lead
Aso. Architect
Jan
2018
Jan 2020
Soft.
Architect
Architecture
Software architecture refers to the high level structures of a software system, the discipline of creating such structures, and the documentation of these structures.
Src: Wikipedia
Architecture
- Choices which are costly to change once implemented
- General understanding of how a system works
Architecture
How we should have built the software system
Architecture
- Simple
- Maintainable
- Scalable
- Fault tolerant
- Cost effective
A NOT scalable codebase
time
features
Difficult to add features overtime
A scalable codebase
time
features
Easier to add features overtime
Factors
- Structure
- Modularity
- Readability
- Repetition for separation
L
I
F
T
Locate the code easily
Identifiable at a glance
Flat structure as much as possible
Try to stay DRY (Don't repeat yourself)
Credit: John Papa
L
Locate the code easily
Which is the controller layer?
Which is the service layer?
WTH is Elixir?
L
Locate the code easily
No need to say anything here ...
Modularity
Break it when you think its doing too much or an entirely different thing
Read Code > Write Code
Meaningful Names
function mashIt(x, y) {
let retVal = x / y;
return retVal;
}function divide(dividend, divisor) {
let result = dividend / divisor;
return result;
}Descriptive Names
function helper.remove (x, y) {
delete x[y];
return x;
}
function deleteProperty(obj, prop) {
delete obj[prop]
return obj;
}Correct Names
function delCostomer (x, y) {
// some logic
// to delete costomer
}
function deleteCustomer(obj, prop) {
// some logic
// to delete customer
}Readable Functions
- Do exactly one thing
- Don't tend to be huge
- Shouldn't have too much nesting
- Shouldn't have too much complexity
Don't Repeat Yourself
- Don't DRY too much
- Don't DRY too little
Look for patterns to make the right decision instead of searching for the silver bullet
Infrastructure
- Memory Pressure
- Compute Pressure
- Network Pressure
Scale
vertical/up or horizontal/out
Memory: 1 GB
vCPU: 2
Memory: 2 GB
vCPU: 4
Memory: 1 GB
vCPU: 2
up/vertical
out/horizontal
How do I know when to scale ?
Monitor
CloudWatch
X Ray
How to scale out ?
Internet
App Server
Database
Internet
Load balancer
App Servers
DB
Cluster
Can't scale out ?
Look for opportunities to cache frequently accessed data in memory
App Server
Cached data
App Memory
In-memory cache
Internet
App Server
Database
Something is fishy
Internet
Load balancer
App Servers
with in-memory cache
DB
Cluster
Distributed Cache
Internet
Load balancer
App Servers
DB
Cluster
Distributed Cache Servers
Caching as a Service
Redis Cache
ElastiCache
Microservices and containers
Some other time!!
Storage types
- Block
- File
- Object
Storage as a Service
Block Storage
S3
Glacier
File Storage
Page Blob
Block Blob
Files
Storage
Databases
Databases need to scale as well
Vertical/Up
Works the same way
Horizontal/Out
- Sharding
- Replication
Database as a Service
SQL
CosmosDB
RDS
DynamoDB
Database as a Service
Architecture needs to be built for
- Fault tolerance
- Disaster recovery
0 downtime deployments
Blue/Green Deployments
Managed Deployment Services
AKS
AppService
EBS
ECS
EKS
Architecture in diagrams
- Knowledge Wiki
- Component
- Infrastructure
- Database
10Pearls Wiki
Login with your COGS credentials
Analyze wants against needs
Not every project needs cross-region replication
Architecture and Scalability
By Mohammad Umair Khan
