https://comparison.metacog.co.kr/tags/normalization/Recent content in Normalization on IT ComparisonHugoen-usSun, 06 Sep 2026 09:38:11 +0900https://comparison.metacog.co.kr/posts/2026-09-06-normalization-vs-denormalization-split-tables-vs-duplicated/Sun, 06 Sep 2026 09:38:11 +0900https://comparison.metacog.co.kr/posts/2026-09-06-normalization-vs-denormalization-split-tables-vs-duplicated/<h2 id="overview">Overview</h2> <p>Normalization organizes data into separate, related tables to eliminate redundancy and protect <strong class="kw">integrity</strong>, while denormalization intentionally merges and duplicates data to boost <strong class="kw">read speed</strong>. The right choice depends on whether your workload is dominated by frequent writes or by heavy, complex reads.</p> <h2 id="comparison-diagram">Comparison Diagram</h2> <div class="compare-diagram"> <svg viewBox="0 0 640 360" xmlns="http://www.w3.org/2000/svg"><text x="160" y="36" text-anchor="middle" font-size="18" style="fill:var(--primary)">Normalization</text><text x="480" y="36" text-anchor="middle" font-size="18" style="fill:var(--primary)">Denormalization</text><rect x="40" y="70" width="180" height="56" rx="4" style="fill:var(--compare-a-soft);stroke:var(--compare-a)" stroke-width="1.5"/><text x="130" y="92" text-anchor="middle" font-size="12" style="fill:var(--content)">Users</text><text x="130" y="108" text-anchor="middle" font-size="10" style="fill:var(--secondary)">id, name</text><rect x="40" y="156" width="180" height="56" rx="4" style="fill:var(--compare-a-soft);stroke:var(--compare-a)" stroke-width="1.5"/><text x="130" y="178" text-anchor="middle" font-size="12" style="fill:var(--content)">Orders</text><text x="130" y="194" text-anchor="middle" font-size="10" style="fill:var(--secondary)">id, user_id, product_id</text><rect x="40" y="242" width="180" height="56" rx="4" style="fill:var(--compare-a-soft);stroke:var(--compare-a)" stroke-width="1.5"/><text x="130" y="264" text-anchor="middle" font-size="12" style="fill:var(--content)">Products</text><text x="130" y="280" text-anchor="middle" font-size="10" style="fill:var(--secondary)">id, name, price</text><line x1="130" y1="126" x2="130" y2="156" style="stroke:var(--compare-a)" stroke-width="1.5"/><path d="M 220 184 C 260 184 260 270 220 270" fill="none" style="stroke:var(--compare-a)" stroke-width="1.5"/><text x="130" y="316" text-anchor="middle" font-size="11" style="fill:var(--secondary)">3 linked tables, zero duplication</text><rect x="360" y="70" width="240" height="228" rx="4" style="fill:var(--compare-b-soft);stroke:var(--compare-b)" stroke-width="1.5"/><line x1="360" y1="108" x2="600" y2="108" style="stroke:var(--compare-b)" stroke-width="1"/><line x1="360" y1="146" x2="600" y2="146" style="stroke:var(--compare-b)" stroke-width="1"/><line x1="360" y1="184" x2="600" y2="184" style="stroke:var(--compare-b)" stroke-width="1"/><line x1="360" y1="222" x2="600" y2="222" style="stroke:var(--compare-b)" stroke-width="1"/><line x1="360" y1="260" x2="600" y2="260" style="stroke:var(--compare-b)" stroke-width="1"/><text x="375" y="92" font-size="11" style="fill:var(--primary)">order_id | customer | product</text><text x="375" y="130" font-size="11" style="fill:var(--content)">101 | Alice | Widget</text><text x="375" y="168" font-size="11" style="fill:var(--content)">102 | Alice | Gadget</text><text x="375" y="206" font-size="11" style="fill:var(--content)">103 | Bob | Widget</text><text x="375" y="244" font-size="11" style="fill:var(--content)">104 | Bob | Gizmo</text><text x="480" y="316" text-anchor="middle" font-size="11" style="fill:var(--secondary)">1 wide table, repeated values</text></svg> </div> <h2 id="comparison-table">Comparison Table</h2> <table> <thead> <tr> <th>Aspect</th> <th>Normalization</th> <th>Denormalization</th> </tr> </thead> <tbody> <tr> <td>Design goal</td> <td>Eliminate redundancy by decomposing data into logical entities</td> <td>Optimize for fast retrieval by pre-combining related data</td> </tr> <tr> <td>Table structure</td> <td>Many narrow, related tables linked by foreign keys</td> <td>Fewer, wider tables that embed related data directly</td> </tr> <tr> <td>Data redundancy</td> <td>Minimal; each fact stored in exactly one place</td> <td>Deliberate; the same fact may appear in many rows</td> </tr> <tr> <td>Write operations</td> <td>Single-row updates ripple correctly since data lives once</td> <td>Updates must touch every duplicated copy or drift occurs</td> </tr> <tr> <td>Read operations</td> <td>Requires assembling data from multiple tables</td> <td>Data is already co-located, so reads are direct</td> </tr> <tr> <td>Joins needed</td> <td>Frequent, often multi-table joins for common queries</td> <td>Rare or none, since data is flattened in advance</td> </tr> <tr> <td>Data integrity risk</td> <td>Low; constraints enforce a single source of truth</td> <td>Higher; duplicate copies can become inconsistent</td> </tr> <tr> <td>Storage requirements</td> <td>Compact, no duplicated values</td> <td>Larger footprint due to stored redundancy</td> </tr> </tbody> </table> <h2 id="key-differences">Key Differences</h2> <ul> <li>Normalization removes <strong class="kw">redundancy</strong> by splitting data into related tables; denormalization reintroduces it deliberately for speed</li> <li>Normalized schemas need more <strong class="kw">joins</strong> at read time, while denormalized ones avoid them by pre-joining data</li> <li>Denormalization trades update simplicity for risk of <strong class="kw">anomalies</strong> when duplicated copies fall out of sync</li> <li>Normalization favors <strong class="kw">write-heavy</strong> transactional workloads; denormalization favors <strong class="kw">read-heavy</strong> analytical ones</li> <li>Storage cost is lower under normalization but query complexity is lower under denormalization</li> </ul> <h2 id="when-to-use-each">When to Use Each</h2> <p><strong>Normalization</strong></p>