Software Deployment
aur Monitoring

Sabse pehle ek word samajh lo: Environment = wo system (ya systems ka set) jo ek given software application ko execute karta hai. Developer code likhta hai, environment usse build karke executable banata hai jise log access karte hain.

Char types hote hain. Lecture mein ek important line hai: development, testing aur production toh usually har team ke paas hote hi hain — staging optional-ish hai par best practice hai.

Staging kyu chahiye — 3 solid reasons

1. Deployment mein bahut saari activities aur configurations hoti hain — libraries install, packages configure, scripts run (jaise DB initialize). Agar ye sab seedha production pe kiya aur kuch galat ho gaya, toh live system crash ho sakta hai aur real users affect honge. Staging mein pehle test karo, phir wahi steps production pe repeat karo (kyunki staging production ki replica hai).
2. New features ka preview — management, sales/marketing teams, ya kuch selected potential clients ko naya feature dikhana ho, par sabko rollout nahi karna — staging pe dikha do.
3. Performance / stress testing — ye local machine ya testing environment pe theek se nahi ho sakti, live-jaisa environment chahiye. Aur production pe karoge toh actual users disturb honge. Isliye staging perfect jagah hai.

Staging se production tak code push karne ke tareeke. Teen strategies padhni hain: Blue/Green, Canary, aur Versioned.

Blue/Green Deployment (Staged Deployment)

New version ke liye ek alag naya production environment banao, current wale ko chhue bina. Blue = abhi ka live production. Green = naya version.

• Sufficient integration aur baaki testing ke baad, router saare users ko blue se green pe reroute kar deta hai. Ab saari incoming requests green pe jaati hain, blue hide ho jata hai.
• Bada advantage: rollback bahut easy hai. Kuch bhi crash hua ya error aaya toh users ko wapas blue pe reroute kar do — blue toh chal hi raha tha.
• Agli update ke liye ab staging se code blue pe push hoga aur switch hoga — matlab blue aur green regularly cycle karte rehte hain real aur previous version ke beech.

Canary Deployment (Phased / Incremental Rollout)

Blue/Green jaisa hi, par ek go mein sabko nahi — users ke chhote subset ko dheere-dheere naya version diya jata hai.

• Users kaise select karein? Randomly, ya profile / demographics ke basis pe, ya region ke basis pe (geographically distributed servers — pehle ek region, sab theek toh aage badho).
• Drawback: ek saath multiple instances manage karne padte hain — kuch users purane version pe, kuch naye pe.

Versioned Deployment

Users ko khud version choose karne do. Saare versions alive rakhe jaate hain. User update kar le toh use new version pe route kar do, warna wo purane pe hi raha.

• Classic example: mobile apps (Paytm, Instagram) — "update available" notification aata hai, aapki marzi.
• Drawback: multiple versions maintain karne padte hain. Isiliye companies bolti hain "in tarikh ke baad in versions ka support band".

Hosting = wo infrastructure jo applications ko host/deploy karne ke liye chahiye. Web apps usually server systems pe host hoti hain, clients wahan se service access karte hain.

Teen hosting options — kaun sa layer kaun sambhalta hai

Neeche har option mein 4 layers hain: Server Hardware → Operating System → Web Server → Web Application. Jitne zyada layer aapke, utna control; jitne kam, utni asaani.

1. Bare Metal Servers

• Aap khud actual server hardware kharidte ho: CPU, motherboard, RAM, disk — sab high configuration mein. Phir OS, web server aur application khud set up karte ho.
• Desktop se alag forms: rack servers, blade servers — compact, taaki ek rack/stack mein kai servers aa jaayen.
• Advantage: highest performance, kyunki server sirf aapka hai, kisi aur ke saath share nahi.
• Disadvantages: most expensive upfront (server lakhon rupees ke aate hain), aur setup mein bahut time+effort, plus ongoing maintenance ka kharcha (part kharab hua toh repair aapka sar-dard).

2. Infrastructure-as-a-Service (IaaS)

• Provider server hardware kharidta hai, us pe OS configure karta hai, aur infrastructure ka ek hissa aapko fees pe deta hai.
• Lecture ka exact example: 8-core processor + 32 GB RAM ko 8 parts mein slice kiya jata hai. Har part ek Virtual Machine (VM) — matlab har VM = 1 core + 4 GB memory. Us VM pe aapka exclusive access.
• VM pe aap code copy karte ho, web server aur database set karte ho. Underlying hardware ki tension nahi.
• Examples: DigitalOcean, AWS, Linode.
• Advantages: sasta (monthly/yearly cost server kharidne se kaafi kam), maintenance overhead nahi.
• Disadvantages: har IaaS ka apna set of configurations hota hai jisse adjust karna padta hai; aur infrastructure doosron ke saath shared hai, isliye performance apne exclusive server jitni nahi hogi.

3. Platform-as-a-Service (PaaS)

• Infrastructure ke upar software layer bhi deta hai jispe web app deploy ho jaati hai. Web server aur database automatically set up ho jaate hain — aap sirf configuration options dete ho.
• Examples: Heroku, Google App Engine.
• Advantage: deploy karna bahut easy — code ready hai toh PaaS baaki sab sambhal leta hai.
• Disadvantage: lack of control — underlying OS kya hai, hardware kya hai, configuration kaisi hai, pata hi nahi chalta.

IaaS Tutorial — DigitalOcean

DigitalOcean humein hardware + operating system deta hai apni app host karne ke liye.

Droplet banane ke steps (video ke hisaab se):

1. Account banao / login karo, aur ek project create karo
2. Region chuno — video mein Bangalore data center, kyunki wo humare closest hai
3. OS aur version chuno
4. Droplet type chuno — simple apps ke liye basic droplet kaafi hai; CPU options regular / premium Intel / AMD; shared CPU vs dedicated CPU ka farq hota hai. Blogs, discussion forums, chhote projects ke liye basic theek. (Ek range hoti hai — upar 256 GB RAM, 4.69 TB storage tak; minimum lagbhag $6/month.) Extra storage bhi add kar sakte ho.
5. Authentication method chuno — SSH key (terminal se connect) ya password. Monitoring aur backups enable kar sakte ho.
6. Droplet ban gaya → usi credentials se login karo → GitHub se code checkout karo, web server + database server set karo, application run karo — bilkul waise hi jaise local system pe karte ho.

PaaS Tutorial — Heroku (Flask app)

Ek simple Flask app (ek route, "hello world") ko Heroku pe deploy karne ka poora flow:

heroku login          # browser khulega, login karo
git init              # project mein git repo initialize
heroku git:remote -a <app-name>   # heroku ko remote repo bana do
git add Procfile app.py requirements.txt
git commit -m "initial commit"
git push heroku master   # push + deployment dono

Code change karna ho? Normal git process: file badlo → git add → git commit → git push heroku master. Har push pe wo commit bhi karega aur build+deploy bhi. Heroku dashboard pe build succeeded dikh jayega; metrics ke liye credit card chahiye; extra resources (data stores, caching, CMS) bhi add kar sakte ho. Google App Engine bhi similar process follow karta hai.

Deployment kitna important aur time-consuming hai ye ab tak samajh aa gaya — multiple logon ka code integrate karo, sufficiently test karo, aur bina bugs ke live environment pe sahi se deploy karo. CI = integration aur deployment process ko automate karna.

7 Best Practices for Continuous Integration

Ye exam mein list ke roop mein poochha jata hai — dohra lo.

CI Server examples

JenkinsCruiseControl — CruiseControl ka dashboard builds ki list dikhata hai: kaunse pass/fail hue, kitni duration lagi, kisne modifications kiye, waghera.

Build tools alag hain: AntGradleBuilder — inhe CI servers ke saath mat mila dena.

Benefits of CI

• Deployment time kam ho jata hai — build, test, deploy sab automated hai, human intervention wale steps minimum ho jate hain.
• Release date pe last-minute confusion aur rush nahi hoti — agar developers hafton/mahinon tak commit na karein aur deadline pe ek saath commit karein, toh bahut conflicts aayenge. CI unhe roz mainline pe commit karne ke liye mandate karta hai, jisse conflicts kam.
• Users ke liye faayda — initial prototype dekh sakte hain; sales/management team ko demo dena ho toh staging server se turant chal jaata hai, kyunki CI server staging ko hamesha updated build deta rehta hai.
• Developers ke liye faayda — frequent commits ki wajah se zyada modular aur efficient code likhne ki aadat padti hai (chhote pieces mein code + commit).

Drawback of CI

Sirf ek: processes set up karne ka initial effort — integration server lagana, logon ko train karna, sab mein time lagta hai. Par zyadatar cases mein advantages drawbacks se kahin zyada hain.

Code deploy ho gaya — ab performance acceptable standards pe honi chahiye. Teen bade tareeke: Caching, Task Queues, aur pages ko jaldi load/render karna.

1. Caching

Common operations ke results ko memory mein store kar lena, taaki har baar database se na laana pade. Memory se lana zyada fast hai aur servers pe load kam hota hai. Common queries agar baar-baar chalti hain aur DB mein koi change nahi hua, toh memory se serve kar do.

Tools: MemcachedRedis

2. Task Queues

Application badi hoti hai toh bahut saare resource aur time consuming tasks aate hain — hazaron users ko email/notification bhejna, bulk database records create/update karna (jaise user delete karo toh uska saara associated data delete karo). Ye tasks turant execute karna zaroori nahi — user ko bata do ki ho jayega aur baad mein notification bhej do.

Iske liye asynchronous work queue use hoti hai. Asynchronous isliye kyunki ye normal HTTP request–response cycle ke bahar execute hoti hai.

Kyu zaroori hai? Normal cycle mein: request → server process kare → email bheje → response de. Par hazaron users ko email bhejne ke liye user ko itni der wait nahi karwa sakte. Isliye task queue mein daal do.

3. Loading & Rendering Pages Faster

Web apps mein bahut saara static content hota hai — HTML, CSS, JavaScript, images. Inka jaldi load hona bahut zaroori hai.

Kyu? Kyunki evidence hai: mobile delays se hone wala stress level horror movie dekhne ke barabar hota hai (Ericsson study wali famous slide). Users bas kuch seconds extra mein hi uncomfortable aur agitated ho jaate hain.

Tareeka A — Minify your code

Minify matlab code se saare spaces, tabs, extra white-space hata dena (aur parameter names tak chhote kar dena). Slide wale simple Fahrenheit→Celsius JavaScript function mein hi ~41% ki saving hui — aur ye toh ek chhota function tha. Poori site pe ye add up hoke bada farq banata hai.

function ftoc(f) {
    var celsius = (f - 32) * 5 / 9;
    return celsius;
}

— minify dabao —

Tareeka B — Compress static pages

Static pages ko compress karo. Tools: BrotliGzip

Testing aur staging mein sab theek tha, phir bhi live environment mein issues aa sakte hain. Isliye deploy ke baad constant monitoring zaroori hai.

Kya-kya monitor karte hain

In sab ke basis pe development aur operations teams respond karke problems fix karti hain.

Monitoring Tools — Report generating tools

Inhe URL do, ye page resources analyze karke optimization suggestions dete hain, SEO aur accessibility metrics check karte hain, aur ek performance score calculate karte hain.

Example: Lighthouse — Google ki service, Chrome extension ke roop mein install hoti hai. Speed index jaise kai metrics deti hai.

Monitoring to understand Customer Behaviour

Analytics Tools

Kaam kaise karte hain: site ke har page mein JavaScript ka chhota piece embed kar do. Ismein tracking code hota hai jo page load hote hi information analytics tool ko bhej deta hai.

Example: Google Analytics — har interaction ke baad ke metrics deta hai, kahan drop-off ho raha hai wo bhi. Bas account sign up karo aur proper tracking code website mein daal do.

Monitoring for A/B Testing

Version A mein kuch issues hain (maan lo abandonment zyada hai). Toh Version B banao — naya workflow ya better UI. Ab canary deployment (incremental rollout) karo: kuch users ko A, kuch ko B.

Dono versions monitor karo aur dekho kaun better hai. "Better" ka matlab ho sakta hai: zyada traffic, zyada dwell time, zyada sales, kam abandonment.

Quiz se 10 minute pehle bas ye padh lo.

Neeche 24 questions hain, IITM graded assignment ke pattern mein (MCQ + MSQ). Option pe click karo → turant sahi/galat + explanation. Sab lecture content se hi bane hain.