System Development and New Technologies

The waterfall model is a development model that proceeds through the phases in order from upstream to downstream, like a waterfall, finalizing each phase's deliverables before moving on. Building in short iterations is agile development, firming up requirements with prototypes is the prototyping model, and combining existing services describes a mashup.

The prototyping model aims to build a prototype early and let users try it, resolving ambiguities in requirements and gaps in understanding between the client and developers at an early stage. Reducing lines of code is not the purpose, and it does not allow omitting tests or eliminating maintenance.

Agile development is an approach that accumulates working software with each short iteration and responds flexibly on the premise that specifications will change; Scrum and XP are representative methods. Finalizing all requirements before developing in one go, waiting for complete documents before starting, and involving users only at the end are close to the waterfall approach and are not agile characteristics.

A sprint is the short, fixed period that forms the unit of development in Scrum; each period repeats the cycle of planning, developing, producing working software, and holding a retrospective. It is not a term for the requirements definition phase or the acceptance test period, nor a period devoted only to bug fixing.

The waterfall model proceeds on the premise of not going back, so when requirement or design errors surface in the downstream testing phase, the cost of returning upstream to redo work is very large. Since each phase finalizes deliverables such as specifications and design documents, tracking progress is actually a strength, and the model suits large developments. Nor is it unable to document requirements.

The incremental model divides the system into functional units and accumulates development and releases in stages; like iterative development models, it has the advantage that some functions can be used early. Waterfall develops the whole in one sequential pass, structured analysis is an analysis method using DFDs and the like, and a mashup is a combination of existing services; none of them fits.

Development basically starts with requirements analysis, which organizes users' needs, then proceeds to external design of the parts visible to users, internal design of the internal implementation, programming, and testing. Orders in which design precedes requirements analysis, or testing precedes programming, cannot hold given the dependencies between deliverables.

External design covers the parts visible to users and the outside — screen specifications, report specifications, and interfaces with other systems. Deciding internal algorithms and physical storage layouts belongs to internal and program design, and verifying individual modules is the work of unit testing.

Unit testing is the first stage of the testing process, checking module by module that each works according to its detailed design. Combining modules to check interfaces is integration testing, the client's acceptance judgment is acceptance testing, and checking performance under heavy load describes load and performance testing.

Testing builds up from the smallest units: first unit tests per module, then integration tests combining the modules, then the system test checking the whole system, and finally the acceptance test by the client and users. Checking the whole or accepting first would run higher-level tests with component defects still mixed in, which is unreasonable.

Acceptance testing is the final-stage test in which the client and users examine the delivered system from a business perspective and judge whether to accept it. Checking internal logic is developer-run unit testing (white-box testing), checking without execution is static testing, and checking module connections is integration testing; none of them describes acceptance testing.

Black-box testing treats the program's interior as a "black box" and checks only whether the input-output relationship matches the specification. White-box testing focuses on covering internal branches and paths, static testing checks through reviews without executing the program, and regression testing checks that fixes have not broken anything else.

Checking whether the system withstands high load — many simultaneous accesses or large data volumes — and keeps its performance is load testing, carried out at the system test stage together with performance tests of response times and the like. Unit testing checks individual modules, static testing checks without execution, and white-box testing is a technique focused on internal structure.

The requirements definition document compiles, as the outcome of requirements analysis, the functions, performance, and constraints the system must deliver, and is agreed between client and developer, serving as the baseline for subsequent design, testing, and acceptance. The operation manual is a how-to document for users, the unit test report records test results, and the network diagram is a design drawing; none is the agreed baseline document.

A WBS is a technique that hierarchically decomposes and structures the project's deliverables and the work needed for them, preventing omissions and clarifying the units for estimation and assignment. Showing schedules as horizontal bars is a Gantt chart, earned-value management is EVM, and evaluating risks describes risk analysis.

A Gantt chart shows each task's planned start and end as horizontal bars and is widely used to visualize schedules and progress. A data flow diagram shows the flow of data, an E-R diagram the relationships between entities, and a use case diagram the interactions between the system and its users; none is a scheduling chart.

There are 2 paths: A→B→D→F (3 + 5 + 4 + 2 = 14 days) and A→C→E→F (3 + 4 + 6 + 2 = 15 days). Since F must wait for both paths to finish, the minimum overall duration is the longer one, 15 days. This longest path, A→C→E→F, is the critical path. 14 days looks only at the A→B→D→F side, and 12 and 13 days match no path total.

There are 3 paths: A→B→D→F (2 + 4 + 3 + 1 = 10 days), A→C→D→F (2 + 6 + 3 + 1 = 12 days), and A→C→E→F (2 + 6 + 4 + 1 = 13 days). The longest, A→C→E→F at 13 days, is the critical path. 10 and 12 days are the totals of the other paths, and 14 days matches none. Any delay in the critical path tasks (A, C, E, F) delays the whole project by the same amount.

The function point method scores the number and complexity of functions — inputs, outputs, inquiries, internal files, and so on — to estimate software size, and can be applied early in development without depending on lines of code. LOC is a different measure based on line counts, the Delphi method converges expert opinions through repeated surveys, and the KJ method is an ideation technique that organizes and groups information on cards.

Risk management proceeds in the order of risk identification, which surfaces the risks; risk analysis (qualitative and quantitative), which evaluates probability and impact; risk response planning, which decides countermeasures; and risk monitoring and control, which tracks the situation during execution. Planning responses or monitoring before identification cannot work, since the risks in question are not yet determined.

Specification changes follow the change control process: the impact on schedule, cost, and quality is assessed, and the change control board (CCB) decides approval or rejection. Unconditional acceptance invites delays and quality decline, while refusing everything loses even necessary changes. On-the-spot fixes by individual staff create inconsistencies with the configuration and documents and destroy control.

Stakeholders are the collective term for individuals and organizations with an interest in the project's outcome — the client, users, developers, management, and so on — and grasping and reconciling their expectations is an important project management activity. The schedule chart is a Gantt chart, the hierarchical decomposition is the WBS, and the list of risks corresponds to the risk register.

An SLA is an agreement between the provider and users of a service on quality levels such as availability, failure recovery time, and support hours, and serves as the evaluation baseline for the operating service. Protecting copyright is a license agreement, the work breakdown chart is a WBS, and borrowing equipment is a lease; none of them is an SLA.

ITIL (Information Technology Infrastructure Library) is a body of documents systematizing best practices for operating and managing IT services, widely consulted as a service management framework covering incident management, change management, the service desk, and more. Programming language standards, normalization theory, and wireless LAN standards belong to other fields.

Availability is calculated as MTBF ÷ (MTBF + MTTR). 570 ÷ (570 + 30) = 570 ÷ 600 = 0.95. MTBF is the mean operating time between failures, and MTTR is the mean time to repair. 0.90 and 0.98 do not match even if numerator and denominator are mixed up, and 0.93 fits no calculation.

Capacity management forecasts future usage — data volumes, access counts — and secures and expands server, storage, and network capacity in a planned way, preventing performance degradation and capacity shortages before they happen. Contact structures for failures belong to incident and emergency management, password management to user management, and license counts to software asset management.

Server virtualization is a technology in which virtualization software runs multiple virtual machines on 1 physical server, improving resource utilization through server consolidation and providing flexibility in reconfiguration and recovery. Physical distribution and smaller chassis are hardware matters, and paper storage has nothing to do with virtualization.

SaaS (Software as a Service) delivers the application itself as a service over the network, letting users work without installing the software. IaaS provides infrastructure such as servers, and PaaS provides everything up to the application runtime environment. On-premises means owning and operating the equipment within one's own organization, which is not a cloud delivery model.

Society 5.0 is a vision of a society that gathers real-world data into cyberspace through IoT and the like, analyzes it, and returns the results to the real world — a high-level fusion of cyberspace and physical space that pursues economic growth and solutions to social problems together. Overlaying information on reality is AR, dividing computation is distributed processing, and the project management body of knowledge describes PMBOK.

Supervised learning gives the model data with correct labels and trains it to predict the answer from the input; it is used for image classification and numerical prediction. Finding structure or groups in data without given answers is unsupervised learning. Training people is not machine learning, and generating random outputs has nothing to do with learning.