Workload

Performance tuning & optimization for Oracle → BigQuery

Oracle performance habits (indexes, optimizer hints, row-by-row PL/SQL, and transactional access patterns) don’t translate. We tune BigQuery layout, queries, and capacity so pruning works, bytes scanned stays stable, and dashboard refresh SLAs hold.

Book assessment Explore workloads

Quick answer

Oracle performance habits (indexes, optimizer hints, row-by-row PL/SQL, and transactional access patterns) don’t translate. We tune BigQuery layout, queries, and capacity so pruning works, bytes scanned stays stable, and dashboard refresh SLAs hold.

Back to pair page

Context

Why this breaks

Oracle workloads are frequently tuned around an OLTP/warehouse database mindset: indexes, optimizer hints, and row-by-row patterns that assume selective access. After migration, teams keep Oracle-era query shapes and expect similar behavior—then BigQuery costs spike because the cost model is dominated by bytes scanned and pruning posture, not indexes. Performance also drifts when incremental applies (MERGE/upserts) touch too much history and when concurrency isn’t planned. Common post-cutover symptoms:

  • Queries scan large tables because filters don’t align to partitions - Join-heavy reports reshuffle large datasets; refresh SLAs slip - Incremental loads and MERGEs scan full targets due to missing scope boundaries - Spend becomes unpredictable because there are no baselines or regression gates - Peak BI usage causes slot contention and tail latency Optimization replaces Oracle’s index/optimizer playbook with BigQuery-native pruning, layout, and governance.

Approach

How conversion works

  • Baseline top workloads: identify the most expensive and most business-critical queries/pipelines (reports, marts, incremental loads). - Diagnose root causes: scan bytes, join patterns, partition pruning, repeated transforms, and MERGE scope. - Tune table layout: partitioning and clustering aligned to access paths and refresh windows. - Rewrite for pruning and reuse: pruning-first filters, pre-aggregation/materializations, and elimination of row-by-row patterns. - Capacity & cost governance: on-demand vs reservations posture, concurrency controls, and guardrails for expensive queries. - Regression gates: store baselines and enforce thresholds so improvements persist.

Coverage

Supported constructs

Representative tuning levers we apply for Oracle → BigQuery workloads.

SourceTargetNotes
Index-driven performance assumptionsPartitioning + clustering aligned to access pathsReplace index selection with pruning-first layout decisions.
Join-heavy reporting queriesPruning-aware rewrites + pre-aggregation/materializationsReduce scan bytes and stabilize dashboard refresh.
MERGE/upsert-based loadsPartition-scoped MERGE and bounded apply windowsAvoid full-target scans and unpredictable runtime.
Row-by-row procedural patternsSet-based SQL refactorsRemove procedural cost cliffs in BigQuery.
Peak user concurrencyReservations/slots + concurrency policiesStabilize SLAs under BI refresh spikes.
Ad-hoc expensive queriesGovernance guardrails + cost controlsPrevent scan blowups and surprise bills.

Compare

How workload changes

TopicOracleBigQueryNotes
Primary cost driverDB capacity + selective access via indexesBytes scanned + slot timePruning and query shape dominate spend.
Tuning focusIndexes, hints, and optimizer behaviorPartitioning/clustering + pruning-first SQLLayout decisions become first-class levers.
Incremental applyTransactional updates + constraintsBounded MERGE/apply with explicit windowsCorrectness and cost depend on apply window design.
Concurrency planningDB sessions and resource mgmtSlots/reservations + concurrency policiesPeak BI usage needs explicit capacity posture.

Examples

Examples

Illustrative BigQuery optimization patterns after Oracle migration: enforce pruning, pre-aggregate for BI, scope MERGEs, and store baselines for regression gates.

01_pruning_first.sql
-- Pruning-first query shape (fact table partitioned by DATE(event_ts))
SELECT
  region,
  SUM(revenue) AS rev
FROM `proj.mart.fact_orders`
WHERE DATE(event_ts) BETWEEN @start_date AND @end_date
GROUP BY 1;
02_preaggregate_for_bi.sql
-- Pre-aggregate for repeated BI scans
CREATE OR REPLACE TABLE `proj.mart.orders_daily`
PARTITION BY d
CLUSTER BY region AS
SELECT
  DATE(event_ts) AS d,
  region,
  SUM(revenue) AS rev,
  COUNT(*) AS orders
FROM `proj.mart.fact_orders`
WHERE DATE(event_ts) BETWEEN @start_date AND @end_date
GROUP BY 1,2;
03_merge_scope.sql
-- Scope MERGE to affected partitions to avoid full-table scans
DECLARE min_d DATE DEFAULT (SELECT MIN(DATE(updated_at)) FROM `proj.ds.stg_orders`);
DECLARE max_d DATE DEFAULT (SELECT MAX(DATE(updated_at)) FROM `proj.ds.stg_orders`);

MERGE `proj.mart.fact_orders` t
USING `proj.ds.stg_orders` s
ON t.id = s.id
AND DATE(t.updated_at) BETWEEN min_d AND max_d
WHEN MATCHED THEN UPDATE SET t.status = s.status, t.amount = s.amount, t.updated_at = s.updated_at
WHEN NOT MATCHED THEN INSERT (id, status, amount, updated_at) VALUES (s.id, s.status, s.amount, s.updated_at);
04_perf_baselines.sql
-- Store performance baselines for regression gates (simplified)
CREATE TABLE IF NOT EXISTS `proj.control.perf_baselines` (
  artifact_id STRING,
  captured_at TIMESTAMP,
  runtime_ms INT64,
  bytes_scanned INT64,
  slot_ms INT64
);

Workload Assessment

Replace index tuning with BigQuery-native pruning

We identify your highest-cost migrated workloads, tune pruning and table layout, and deliver before/after evidence with regression thresholds—so performance improves and stays stable.

Book assessment

Avoid

Common pitfalls

  • Index-thinking carryover: expecting selective index access; BigQuery needs pruning and layout-to-filter alignment.
  • Hints carried over mentally: optimizer hints removed, but query still assumes the same plan.
  • Partitioning after the fact: migrating tables without aligning partitions to common filters and refresh windows.
  • Clustering by folklore: clustering keys chosen without evidence from predicates/join keys.
  • Unbounded MERGE: apply touches too much history and scans full targets.
  • Row-by-row procedural remnants: leftover procedural patterns cause cost cliffs.
  • Ignoring concurrency: BI refresh spikes overwhelm slots/reservations and create tail latency.
  • No regression gates: performance improves once, then regresses silently.

Proof

Validation approach

  • Baseline capture: runtime, bytes scanned, slot time, and output row counts for top queries/pipelines.
  • Pruning checks: confirm partition pruning and predicate pushdown on representative parameters.
  • Before/after evidence: demonstrate improvements in runtime and scan bytes; document tradeoffs.
  • Correctness guardrails: golden queries and KPI aggregates ensure tuning doesn’t change semantics.
  • Regression thresholds: define alerts (e.g., +25% bytes scanned or +30% runtime) and enforce via CI or scheduled checks.
  • Operational monitors: dashboards for scan bytes, slot utilization, failures, and refresh SLA adherence.

Execution

Migration steps

A sequence that improves performance while protecting semantics.

  1. 01

    Identify top cost and SLA drivers

    Rank queries and pipelines by bytes scanned, slot time, and business criticality (dashboards, batch windows). Select a tuning backlog with clear owners.

  2. 02

    Create baselines and targets

    Capture current BigQuery job metrics (runtime, scan bytes, slot time) and define improvement targets. Freeze golden outputs so correctness doesn’t regress.

  3. 03

    Tune layout: partitioning and clustering

    Align partitions to common filters and refresh windows. Choose clustering keys based on observed predicates and join keys—not guesses.

  4. 04

    Rewrite for pruning and reuse

    Apply pruning-aware rewrites, reduce reshuffles, scope MERGEs/applies to affected partitions, and pre-aggregate where BI repeatedly scans large facts.

  5. 05

    Capacity posture and governance

    Set reservations/on-demand posture, tune concurrency for BI refresh peaks, and implement guardrails to prevent scan blowups from new queries.

  6. 06

    Add regression gates

    Codify performance thresholds and alerting so future changes don’t reintroduce high scan bytes or missed SLAs. Monitor post-cutover metrics continuously.

FAQ

Frequently asked questions

Why did costs increase after moving from Oracle to BigQuery? +

Most often because Oracle queries assumed selective access via indexes. In BigQuery, bytes scanned is the dominant cost driver, so pruning-first filters and layout alignment (partitioning/clustering) are required to stabilize spend.

How do you keep optimization from changing results? +

We gate tuning with correctness checks: golden queries, KPI aggregates, and edge-cohort diffs. Optimizations only ship when outputs remain within agreed tolerances.

Can you optimize MERGE/upsert pipelines too? +

Yes. We scope MERGEs to affected partitions, design staging boundaries, and validate performance with scan bytes/slot time baselines to prevent unpredictable runtime.

Do you cover reservations and concurrency planning? +

Yes. We recommend a capacity posture (on-demand vs reservations), concurrency controls for BI refresh spikes, and monitoring/guardrails so performance stays stable as usage grows.

Optimization Program

Prevent scan blowups with regression gates

Get an optimization backlog, tuned partitioning/clustering, and performance gates (runtime/bytes/slot thresholds) so future releases don’t reintroduce slow dashboards or high spend.

Book assessment
Book assessment