Software Maintenance & Evolution: Types, Challenges & Strategies

Key Concepts and Definitions

What is Software Maintenance?

Software maintenance is the modification of a software product after delivery to correct faults, improve performance, or adapt to a changed environment. It is not a one-time activity but an ongoing process that can consume 40–80% of a software’s total lifecycle cost.

stateDiagram-v2
    [*] --> Idle
    Idle --> Active: Deployment
    Active --> Corrective: Fault Reported
    Active --> Adaptive: Environment Change
    Active --> Perfective: User Feedback
    Active --> Preventive: Risk Mitigation
    Active --> [*]: Retirement

Types of Software Maintenance

Maintenance is classified into four main types, each addressing different needs:

Why Software Evolves: Drivers of Change

Software does not remain static. Key drivers include:

sequenceDiagram
    participant User
    participant BankSystem
    participant Regulator
    participant Competitor
    User->>BankSystem: Requests loan (new KYC rules)
    Regulator->>BankSystem: Updates compliance API
    Competitor->>BankSystem: Adds AI route optimization
    BankSystem-->>User: Approves loan with new features
    Note right of BankSystem: Drivers: Requirements, Regulations, Competition

1. Changing User Requirements

   • Example: Nepal Rastra Bank (NRB) mandates new KYC (Know Your Customer) rules for banks → Nabil Bank must update its loan processing software.

2. Technological Advancements

   • Example: Daraz migrates from monolithic architecture to microservices to handle increased traffic during sales.

3. Legal and Regulatory Changes

   • Example: NTC updates its billing software to comply with new GDPR-like data privacy laws in Nepal.

4. Performance and Security Issues

   • Example: Ncell patches its mobile app after a DDoS attack exposes a flaw in its authentication system.

5. Competitive Pressures

   • Example: Pathao adds AI-driven route optimization to compete with Karma Taxi.

IMAGE: server rack in data center | Hardware infrastructure supporting software maintenance (e.g., patch deployment, backups)

Software Maintenance Process

The process involves planning, implementation, and review in a structured workflow:

flowchart TD
    A["Maintenance Request"] --> B{"Type of Maintenance?"}
    B -->|"Corrective"| C["Debugging & Fix"]
    B -->|"Adaptive"| D["Environment Analysis"]
    B -->|"Perfective"| E["Feature Enhancement"]
    B -->|"Preventive"| F["Refactoring/Redesign"]
    C --> G["Testing"]
    D --> G
    E --> G
    F --> G
    G --> H["Deployment"]
    H --> I["User Feedback/Monitoring"]
    I -->|"Loop"| A

Steps in Detail:

1. Request Analysis

   • Prioritize based on severity, impact, and cost.
   • Example: A bank’s core banking system crash during Diwali season (high priority) vs. a minor UI glitch (low priority).

2. Impact Analysis

   • Assess how changes affect other modules, performance, and security.
   • Example: Updating NEPSE’s trading software may require load testing to handle increased volume.

3. Modification and Testing

   • Use unit, integration, and regression testing.
   • Example: Khalti tests a new two-factor authentication feature with real users before full rollout.

4. Implementation

   • Deploy fixes in stages (e.g., canary releases for YouTube updates).

5. Review and Documentation

   • Update user manuals, API docs, and code comments.
   • Example: Google maintains a public changelog for Android updates.

Software Reengineering and Reverse Engineering

When maintenance becomes too costly, reengineering (restructuring existing software) or reverse engineering (analyzing undocumented code) is used.

IMAGE: motherboard with CPU and RAM | Hardware constraints (e.g., legacy systems) often drive software reengineering

Cost of Software Maintenance

Maintenance costs can be high but necessary. A study by IBM found that software maintenance accounts for 60–80% of total lifecycle costs.

Cost Factors:

• Direct Costs: Labor, tools, testing.
• Indirect Costs: Downtime, lost revenue, user dissatisfaction.
• Example: Daraz’s website crash during Dashain sales costs millions in lost sales.

Cost Estimation Models:

1. Function Point Analysis (FPA)
   • Measures maintenance effort based on function points (inputs, outputs, queries).
   • Example: A bank’s loan processing module with 500 function points may cost $50,000/year to maintain.

2. COCOMO Model
   • Estimates effort based on lines of code (LOC) and complexity.
   • Example: A 100,000 LOC system may require 2 person-years for corrective maintenance.

Tools for Software Maintenance

IMAGE: GitHub interface showing commit history | Version control tools track every change in software evolution

Challenges in Software Maintenance

1. Legacy Systems

   • Example: Nepal’s old land record system runs on 1990s COBOL and lacks modern APIs.

2. Lack of Documentation

   • Example: A bank’s retired developer leaves without documenting a critical module.

3. Changing Technology

   • Example: WhatsApp must constantly update its end-to-end encryption to counter new threats.

4. User Resistance

   • Example: Elderly users of eSewa may reject a new biometric login system.

5. Security Risks

   • Example: Pathao’s old SQL queries are vulnerable to SQL injection attacks.

Strategies for Effective Maintenance

1. Modular Design

   • Example: Google’s microservices architecture allows teams to update YouTube’s recommendation engine independently.

2. Automated Testing

   • Example: Khalti uses CI/CD pipelines to auto-test every commit.

3. Regular Refactoring

   • Example: Facebook spends 20% of dev time on refactoring to reduce technical debt.

4. User Feedback Loops

   • Example: Daraz uses surveys and analytics to prioritize maintenance requests.

5. Training and Knowledge Transfer

   • Example: NTC conducts workshops for engineers maintaining its legacy switching systems.

In the Real World

1. eSewa’s Maintenance Challenges

   • Problem: During Dashain, eSewa’s servers crash due to unexpected transaction spikes.
   • Solution: They implemented auto-scaling in AWS and caching mechanisms to handle load.
   • Lesson: Perfective maintenance (performance tuning) is critical for high-traffic systems.

2. Khalti’s Adaptive Maintenance

   • Problem: Nepal Rastra Bank introduces new KYC rules.
   • Solution: Khalti’s team spent 3 months updating its identity verification module.
   • Lesson: Adaptive maintenance ensures compliance and avoids legal penalties.

3. NEPSE’s Corrective Maintenance

   • Problem: A bug in NEPSE’s trading software causes false buy/sell orders.
   • Solution: Emergency patch deployed during market hours with rollback plans.
   • Lesson: Corrective maintenance must be fast and reliable to prevent financial losses.

Worked Example: Maintaining a Bank’s Loan Processing System

Scenario: A bank’s loan approval system is slow during peak hours (e.g., Diwali season). The IT team must decide how to fix it.

Step 1: Identify the Maintenance Type

• Problem: Performance bottleneck → Perfective maintenance (enhancement).
• Alternative: If the system crashes → Corrective maintenance.

Step 2: Analyze the Root Cause

• Possible Causes:
  • Inefficient SQL queries.
  • Lack of database indexing.
  • Monolithic architecture slowing down requests.

Step 3: Propose Solutions

Step 4: Choose and Implement

• Best Option: Optimize queries + add caching (balanced cost/benefit).
• Tools Used:
  • SQL Profiler (to find slow queries).
  • Redis (for caching).
  • Load testing (to verify improvements).

Step 5: Monitor and Iterate

• Post-deployment: Track response times and user feedback.
• Feedback: If issues persist, consider microservices in the next phase.

Exam Tip

What Examiners Look For:

1. Clear Classification

   • Always distinguish between corrective, adaptive, perfective, and preventive maintenance with real-world examples.
   • Example: "eSewa’s security patch is corrective, while its new OTP feature is perfective."

2. Cost-Benefit Analysis

   • Exams often ask: "Which maintenance strategy is best for [scenario]?"
   • Structure your answer:
     • Identify the type of maintenance needed.
     • List pros/cons of each option.
     • Justify your choice with cost, risk, and benefit.

3. Tools and Techniques

   • Know at least 2 tools for each maintenance activity (e.g., Git for version control, SonarQube for static analysis).
   • Example: "For reverse engineering, tools like JAD (Joint Application Design) and decompilers are used."

4. Real-World Scenarios

   • Always tie answers to Nepalese or global companies (e.g., NTC’s legacy systems, Khalti’s compliance updates).
   • Example: "NEPSE’s trading system requires high availability, so preventive maintenance (like load testing) is critical."

5. Diagrams and Flowcharts

   • Draw maintenance process flows (like the one above) to visualize steps.
   • Use tables to compare costs, risks, and benefits of different strategies.

Common Pitfalls to Avoid:

• Vague Answers: Instead of "maintenance is important", say "adaptive maintenance ensures Khalti complies with NRB’s new API standards."
• Ignoring Trade-offs: Always discuss cost vs. benefit (e.g., "Refactoring is expensive but reduces technical debt long-term.").
• Overlooking Documentation: Maintenance fails without proper logs and docs—mention this in answers.

Final Checklist for Full Marks:

✅ Define software maintenance and its four types with Nepali examples. ✅ Explain why software evolves (requirements, tech, laws) with real cases. ✅ Describe the maintenance process (request → analysis → modification → testing → deployment). ✅ Compare reengineering vs. reverse engineering with a Nepalese system (e.g., NTC’s old switches). ✅ Discuss cost models (COCOMO, FPA) and tools (Git, SonarQube). ✅ Solve a worked example (e.g., bank loan system) with step-by-step reasoning. ✅ Link to real-world apps (eSewa, Khalti, NEPSE) in every major section.