| Course | BU 455 Big Data Management |
|---|---|
| Unit | Unit 2 |
| Paper type | Conventional alternative analysis |
| Length | About 900 words, 6 pages |
| Format | APA 7 student paper |
| School | Herzing University |
| Program | BS in Business Analytics |
| Updated | October 2026 |
Free sample paper for BU 455 Unit 2
The Ordinary Database Passed Five of Six Tests: A Conventional Alternative Analysis for a Composite Bike-Share Operator, and the Requirement That Broke It First
Student Name
Business, Herzing University
BU 455: Big Data Management
Instructor Name
January 28, 2026
The Ordinary Database Passed Five of Six Tests: A Conventional Alternative Analysis for a Composite Bike-Share Operator, and the Requirement That Broke It First
Testing the Ordinary Option First
The scale assessment for Pedalworks Mobility, the composite bike-share operator in this course, found about 85 gigabytes of raw data a year, peak arrival of 40 to 50 messages a second and a real problem with mixed formats. Before any distributed platform is considered, this analysis tests whether the company's existing kind of system, a single managed PostgreSQL database, could meet its requirements if properly used. It tests the conventional option honestly, identifies which requirement would break it first, and reports what actually failed rather than assuming something would.
Stonebraker and Cetintemel (2005) argued that no single database design suits every workload, and that systems built for transactions and systems built for analysis make different trade-offs. That argument cuts both ways: it warns against forcing analysis onto a transactional database, but it also means the right question is which workloads strain the current system, not whether a bigger system exists.
The Test Setup
A copy of the production database was created on a managed instance with 8 processor cores, 64 gigabytes of memory and 1 terabyte of solid-state storage, the size the company's cloud account already allows without a new contract. One full year of data was loaded: 1.9 million trips, about 184 million telemetry messages stored in a column that holds JSON documents, 163 million station status rows and 104 million app events. Loaded with indexes, the year occupied about 52 gigabytes, which projects to about 260 gigabytes for five years. Six requirements were drawn from the operations, finance and analytics leads, each with a measurable pass condition.
| Requirement | Pass condition | Result | Verdict |
|---|---|---|---|
| R1. Hold five years of data | Fits within 1 TB with room to grow | About 260 GB projected | Pass |
| R2. Accept peak message rate | Sustain 10 times peak, 500 messages a second, for an hour | Sustained 500 a second; write delay under 20 milliseconds | Pass |
| R3. Alert crews to full or empty stations | Within 2 minutes of the status change | 95% of alerts within 4 seconds | Pass |
| R4. Refresh the daily ridership dashboard | Under 1 minute | 12 seconds | Pass |
| R5. Let analysts query a full year of telemetry | Under 2 minutes without slowing operations | 14 minutes 20 seconds; alert delay rose to 48 seconds during the query | Fail |
| R6. Combine telemetry layouts, notes and photos | One query can use all three telemetry layouts | Works through a view that translates the three layouts | Pass, with upkeep |
What Actually Failed
Only one requirement failed, and it failed in a specific way. The analytics lead's test question was how battery capacity has declined for each bike over the year, which requires reading every telemetry message, extracting the battery field from three different layouts and averaging by bike and week. The query read about 184 million rows and took just over 14 minutes. That time alone might be tolerable for an occasional analysis, but the query also used most of the machine's processing capacity while it ran, and the station alerts that normally arrived within 4 seconds slowed to 48 seconds. An analyst's curiosity was degrading the rebalancing crews' tools.
The failure is therefore not that the database is too small. It is that one machine was doing two jobs with different needs, fast small writes and reads for operations and long full scans for analysis, and the second interfered with the first. This matches the distinction Stonebraker and Cetintemel draw between transactional and analytical workloads.
Fixing the Failure Within the Conventional Option
Three changes were tested, all within ordinary PostgreSQL. A read replica, a continuously updated copy of the database, was added for analysts, so long queries run there and cannot slow the primary. Telemetry was partitioned by month, so queries about recent months read only those partitions. And a nightly job builds a summary table holding each bike's average battery reading per day, about 1.2 million rows a year instead of 184 million messages. With those changes, the battery question ran in 3 seconds against the summary table, the full-scan version ran in about 9 minutes on the replica, and station alerts on the primary stayed at 4 seconds throughout. R5 passes.
The replica roughly doubles the database's monthly cost, and the summary table must be maintained as a new job. Both are modest compared with a new platform. Kleppmann (2017) describes separating operational and analytical reads through replicas and derived tables as a standard way to scale a relational system before replacing it.
How the Variety Problem Was Handled
Requirement R6 passed, but it deserves a note, because variety was the dimension the scale assessment found under real strain. The three telemetry layouts were reconciled with a view, a stored query that presents every message in one standard shape, converting the battery reading from volts to percentage where the older layout used volts. PostgreSQL's support for JSON documents, including functions that pull fields out of them, made this possible without restructuring the stored data (PostgreSQL Global Development Group, 2024). Mechanic notes and photo references were linked to bikes and dates through ordinary tables. The cost is upkeep: each new firmware layout requires a change to the view, and someone must notice when a new layout appears. A weekly check that counts messages by layout now flags any layout the view does not recognize, so a new firmware release cannot silently produce missing battery readings.
What Would Break It Next
The conventional option, with a replica, partitions and summary tables, meets every current requirement. The change most likely to break it is the bike maker's proposed firmware update sending positions every 5 seconds instead of 30. Ingestion would still pass, at roughly 300 messages a second at peak against a tested 500, but five years of telemetry would approach 1.2 terabytes, beyond the current storage, and full scans on the replica would take close to an hour. That threshold is the trigger for the storage and format decisions in the next unit. Until then, concluding that nothing needs to change except how the existing database is used is a legitimate and well-supported answer. Pedalworks and its test results are invented for the course.
References
Kleppmann, M. (2017). Designing data-intensive applications: The big ideas behind reliable, scalable, and maintainable systems. O'Reilly Media.
PostgreSQL Global Development Group. (2024). PostgreSQL 16 documentation. https://www.postgresql.org/docs/16/
Stonebraker, M., & Cetintemel, U. (2005). "One size fits all": An idea whose time has come and gone. In 21st International Conference on Data Engineering (pp. 2-11). IEEE. https://doi.org/10.1109/ICDE.2005.1
Reading the BU 455 Unit 2 instructions
In Unit 2, BU 455 asks students to evaluate whether a conventional database or system could meet the case's requirements. Common requirements include stating each requirement with a measurable condition, testing the conventional option properly, identifying which requirement fails first and why, and considering fixes within the conventional approach before recommending anything distributed. Many instructors reassure students that concluding nothing needs to change can score well if the testing is sound. The unit differs from a database design course because the focus is on workload limits rather than modeling. The deliverable is usually a test table and APA 7 discussion, with results reported as measured rather than assumed.
Inside the BU 455 Unit 2 example
The paper begins by explaining why the conventional option is tested first and cites the argument that different workloads suit different designs. The setup section describes the test instance and the year of data loaded. A table lists six requirements with pass conditions, results and verdicts. A section analyzes the one failure and shows it came from two workloads competing on one machine. A section tests three fixes within PostgreSQL and reports the new results and costs. A section on variety explains how a translation view reconciled three telemetry layouts, with a weekly check. The final section names the change that would break the conventional option next and when to revisit.
BU 455 Unit 2 rubric: what full points look like
Conventional alternative rubrics usually reward measurable requirements, a fair test, accurate identification of the first failure, attempts to fix it conventionally and a justified conclusion, with writing and sources. Requirements earn full credit when each has a pass condition. The fairness criterion rewards realistic data volumes and configurations rather than a straw man. Failure identification scores highest when the student explains the mechanism, not just the symptom. Conventional fixes earn points when tested and priced. The conclusion rewards stating clearly whether change is needed and what would trigger it. Instructors often credit students who report that the ordinary option works, since that requires as much evidence as recommending a new platform.
BU 455 Unit 2 help before you submit
Conventional alternative analyses often lose points by dismissing the ordinary option without testing it. Another common problem is testing with a tiny sample that proves nothing about a year of data. Students frequently report that a query was slow without examining whether it disturbed other work. Some jump straight to distributed platforms when a replica or summary table would solve the problem. Others ignore variety, which a relational database can often handle with views. Many never state what would make the conventional option fail later. A test report that names its own expiry date, such as a firmware change that multiplies telemetry, reads as judgment rather than loyalty to the current system. Finally, using your company's real database performance figures can disclose confidential details; test data built for the course show the same reasoning.
Get a BU 455 Unit 2 example written to your instructions
Pass along the Unit 2 instructions, the rubric your BU 455 section posts and the scenario the assignment builds on. We write a custom example that fixes measurable requirements first, gives the conventional design its ordinary remedies, scores both options identically and states which requirement forces a change, if any does. First custom sample free, returned in 24 to 48 hours.
More BU 455 papers and business analytics samples
- BU 455 Unit 1: Scale assessment
- BU 455 Unit 3: Storage and format decisions
- BU 455 Unit 4: Batch and streaming design
- BU 455 Unit 5: Distributed processing exercise
- BU 455 Unit 6: Governance and retention plan
- BU 455 Unit 7: Cost and capability analysis
- BU 455 Unit 8: Architecture recommendation
- BU 407 Unit 2: Expected value comparison
- BU 451 Unit 2: Data import and inspection script
- BU 453 Unit 2: Entity relationship diagram
- BU 454 Unit 2: Chart critique
BU 455 Unit 2 questions, answered
Will I lose marks for concluding that nothing needs to change?
Not if the analysis supports it, and that verdict is often the one the unit was built to elicit. Show the requirements, show the conventional design meeting them, and say what growth would change the answer. What loses marks is reaching that conclusion by not examining the alternative at all, which reads as avoidance rather than as judgment.
How is this different from a database design course?
A design course asks you to model a domain and write queries against it, and it assumes the system fits the problem. This unit asks whether it still does, which means the subject is the boundary rather than the schema. If your draft spends its length on tables and keys, it has answered the other course and left this criterion unaddressed.
Can I use what I know about my company's database performance?
The query times, row counts and hardware in an employer's environment are operational records and stay there, however ordinary they look to you. Argue from the case the assignment supplies and from published documentation you can cite with a date. Your familiarity with what actually slows a system down will show in which requirements you choose to test, and that is where it belongs.
Where can I find a free BU 455 Unit 2 sample paper?
This page publishes the complete BU 455 Unit 2 conventional alternative analysis free, including the six-requirement table, the fixes and margin notes. For your own case, forward the instructions and a custom analysis is drafted free, typically within 48 hours.
Will I lose marks in BU 455 Unit 2 for concluding nothing needs to change?
Not if the testing supports it. The sample concludes the existing database works with a replica and partitions, and backs that with measured results and a named trigger for revisiting.