Concept card

HANA Performance Tuning

As of 2026-07-23

Edited by Cédric-Olivier Mâry, sole editor and publication director — 27 years in the SAP ecosystem. How these pages are produced, and where automation is used.

What is HANA Performance Tuning?

HANA tuning is what separates the architect who reads query plans from the junior who blames 'the data volume' — five levers convert a 5-50 second query into a 100-500ms one.

What it is

HANA Performance Tuning is the engineering discipline of making queries run in 100-500 ms instead of 5-50 seconds. It separates the senior architect (who reads query plans, designs partitions, knows when to flatten a hierarchy) from the junior consultant (who blames "the data volume"). For SAP analytics consultants in the BDC era — where Datasphere sits on HANA Cloud (companion C001) — HANA fluency is the architectural currency that converts modeling decisions into milliseconds, and milliseconds into user adoption.

The five tuning levers in priority order. (1) Column-store fitness — confirm tables are columnar (default, but legacy migrations sometimes keep row-store). 10-100× scan-speed difference. (2) Push-down preservation — query plans show whether HANA optimizer pushed predicates + projections to the storage layer. Script views (companion C004) break push-down; rewrite as SQL. (3) Partition design — partition fact tables by FISCAL_YEAR (2-5M rows/partition optimal). Partition pruning avoids scanning irrelevant data. (4) Statistics + indexing — collect statistics on join columns; create inverted indexes on filter-heavy columns. (5) Materialization vs federation — if a Datasphere view federates 5+ sources, materialize the join result; query latency drops 5-20×.

Why it matters

Script views break push-down silently — rewriting as SQL alone can produce a step-change in latency that no amount of hardware scaling would fix.
Federating 5+ sources in a Datasphere view without materializing the join result costs 5-20x latency — a decision made at modeling time, not a runtime problem to patch later.
A senior spends 30-60 minutes reading the Plan Visualizer's operator tree per slow query and names the exact operator at fault — vague 'it's slow' diagnoses don't survive that discipline.

Key points

Five tuning levers: column-store · push-down · partition · stats/index · materialize-vs-federate.
Plan Visualizer is the diagnostic tool; reading it takes 5-10 min practice.
Push-down preservation: no Script views in AM dependency tree.
Partition design: FISCAL_YEAR + 2-5M rows/partition optimal.
Materialize at 5+ federated sources (5-20× latency improvement).
Calculation Views ≤ 2 levels deep.
Tuning ROI: 100-1000× over adding capacity.
DAC overhead: 3-5ms/rule × N stacked.
HANA Performance Tuning is mastered only when it changes a named buyer decision.
Start with the semantic contract and control model before demonstrating the tool.

Terms used on this page

Column-store: HANA's default storage layout: columns stored separately, optimal for analytical scans. 10-100× faster than row-store.
Push-down: Optimizer pushing predicates + projections to the storage layer. Script views break it.
Plan Visualizer: HANA Studio tool showing operator-tree per query. The diagnostic primary tool.
Partition pruning: Optimizer skipping irrelevant partitions during scan. Requires partition design (FISCAL_YEAR typical).
Materialize-vs-federate: Trade-off: federated multi-source query is slower but data fresh; materialized join is faster but copy overhead.
Cold cache: First-of-day query is slower because data not yet in HANA RAM. Pre-warm hot models on schedule.
Calculation View nesting: Layered HANA views where each level adds optimization complexity. Cap ≤ 2 levels for readability.
DAC overhead: Per-rule query-rewrite cost: 3-5ms per rule per query. Stacked rules add up.

Sources

Full card available to members. What the full card adds: the full decision framework · the SAP vs Snowflake / Databricks / Fabric comparison · the common pitfalls and their fix · the cheat sheet · the architecture schemas · the code blocks.

Open in the app →