ABAP Cloud Performance Optimization: 10 Quick Wins (2025)

kategorie
Performance
Veröffentlicht
autor
Johannes

Performance-Probleme in ABAP Cloud? Diese 10 Quick Wins bringen sofortige Verbesserungen - mit Code-Beispielen.

🎯 Performance Mindset

Golden Rules:

  1. Measure First - Nicht raten, messen!
  2. DB > ABAP - Arbeit in Datenbank pushen
  3. Batch > Loop - Operationen bündeln
  4. Cache Smart - Aber nicht überoptimieren

1. SELECT ohne WHERE vermeiden

❌ Schlecht (lädt ALLES)

SELECT * FROM I_Customer
INTO TABLE @DATA(lt_customers).
LOOP AT lt_customers INTO DATA(ls_customer)
WHERE Country = 'DE'.
" Verarbeitung
ENDLOOP.

Problem: 1 Mio Zeilen geladen, 950k verworfen!

✅ Gut (WHERE in DB)

SELECT Customer, CustomerName, Country
FROM I_Customer
WHERE Country = 'DE'
INTO TABLE @DATA(lt_customers).

Speedup: ~95% (bei 1 Mio Zeilen)


2. SELECT in Loops vermeiden

❌ Schlecht (N+1 Problem)

LOOP AT lt_orders INTO DATA(ls_order).
" ❌ SELECT pro Iteration!
SELECT SINGLE CustomerName
FROM I_Customer
WHERE Customer = @ls_order-customer
INTO @DATA(lv_name).
ENDLOOP.

Bei 1000 Orders: 1000 SELECT-Statements!

✅ Gut (FOR ALL ENTRIES)

" Ein SELECT für alle
SELECT Customer, CustomerName
FROM I_Customer
FOR ALL ENTRIES IN @lt_orders
WHERE Customer = @lt_orders-customer
INTO TABLE @DATA(lt_customers).
" In Memory joinen
LOOP AT lt_orders ASSIGNING FIELD-SYMBOL(<order>).
READ TABLE lt_customers INTO DATA(ls_cust)
WITH KEY customer = <order>-customer.
<order>-customer_name = ls_cust-customername.
ENDLOOP.

Speedup: ~90% (1000 → 1 DB-Call)


3. CDS Views statt ABAP Joins

❌ Schlecht (Join in ABAP)

SELECT Customer, CustomerName FROM I_Customer
INTO TABLE @DATA(lt_customers).
SELECT SalesOrder, Customer, NetValue FROM I_SalesOrder
INTO TABLE @DATA(lt_orders).
" Join in ABAP ❌
LOOP AT lt_orders ASSIGNING <order>.
READ TABLE lt_customers INTO DATA(ls_cust)
WITH KEY customer = <order>-customer.
" ...
ENDLOOP.

✅ Gut (Join in CDS)

define view Z_CustomerOrders
as select from I_Customer
inner join I_SalesOrder
on I_Customer.Customer = I_SalesOrder.Customer
{
key I_Customer.Customer,
I_Customer.CustomerName,
I_SalesOrder.SalesOrder,
I_SalesOrder.NetValue
}
SELECT * FROM Z_CustomerOrders
INTO TABLE @DATA(lt_result).

Speedup: ~80% (DB macht Join effizienter)


4. EML-Operationen batchen

❌ Schlecht (Loop)

LOOP AT lt_book_ids INTO DATA(lv_id).
" ❌ MODIFY pro ID
MODIFY ENTITIES OF zi_book IN LOCAL MODE
ENTITY Book
UPDATE FIELDS ( Status )
WITH VALUE #( ( BookId = lv_id Status = 'F' ) ).
ENDLOOP.
COMMIT ENTITIES.

Bei 100 Büchern: 100× DB-Zugriff!

✅ Gut (Batch)

" Ein MODIFY für alle
MODIFY ENTITIES OF zi_book IN LOCAL MODE
ENTITY Book
UPDATE FIELDS ( Status )
WITH VALUE #( FOR lv_id IN lt_book_ids
( BookId = lv_id Status = 'F' ) )
FAILED DATA(failed).
COMMIT ENTITIES.

Speedup: ~85%


5. Projection statt SELECT *

❌ Schlecht (alle Felder)

SELECT * FROM I_Product
INTO TABLE @DATA(lt_products).

Lädt 50 Felder, nutzt aber nur 3!

✅ Gut (nur benötigte Felder)

SELECT Product, ProductName, Price
FROM I_Product
INTO TABLE @DATA(lt_products).

Speedup: ~60% (weniger Netzwerk-Traffic)


6. Table Buffering aktivieren

Für Master-Daten (selten ändert sich):

@AbapCatalog.buffering.status: #ACTIVE
@AbapCatalog.buffering.type: #FULL
@AbapCatalog.buffering.numberOfKeyFields: 1
define view Z_Country
as select from T005
{
key land1 as Country,
landx as CountryName
}

Speedup: ~95% (Daten aus Memory statt DB)

ABER: Nur für Daten die selten ändern!


7. FILTER statt LOOP + IF

❌ Schlecht

DATA lt_active_customers TYPE TABLE OF ty_customer.
LOOP AT lt_customers INTO DATA(ls_customer)
WHERE status = 'ACTIVE'.
APPEND ls_customer TO lt_active_customers.
ENDLOOP.

✅ Gut (FILTER Expression)

DATA(lt_active) = FILTER #( lt_customers
WHERE status = 'ACTIVE' ).

Speedup: ~30% (weniger Code, optimiert)


8. REDUCE statt LOOP für Aggregationen

❌ Schlecht

DATA lv_total TYPE p.
LOOP AT lt_orders INTO DATA(ls_order).
lv_total = lv_total + ls_order-amount.
ENDLOOP.

✅ Gut (REDUCE)

DATA(lv_total) = REDUCE p( INIT sum = 0
FOR ls_order IN lt_orders
NEXT sum = sum + ls_order-amount ).

Speedup: ~20%


9. Lazy Loading mit Associations

❌ Schlecht (alles laden)

define view Z_CustomerWithOrders
as select from I_Customer
left outer join I_SalesOrder
on I_Customer.Customer = I_SalesOrder.Customer
{
I_Customer.Customer,
I_Customer.CustomerName,
I_SalesOrder.SalesOrder,
I_SalesOrder.NetValue
}

Problem: Auch wenn Orders nicht gebraucht werden, immer Join!

✅ Gut (Association)

define view Z_Customer
as select from I_Customer
association [0..*] to I_SalesOrder as _Orders
on $projection.Customer = _Orders.Customer
{
key Customer,
CustomerName,
_Orders // Nur exposen, kein Join!
}

Verwendung:

" Nur Customer
SELECT Customer, CustomerName
FROM Z_Customer
INTO TABLE @DATA(lt_customers).
" Mit Orders (nur wenn nötig!)
SELECT Customer, \_Orders-SalesOrder
FROM Z_Customer
INTO TABLE @DATA(lt_with_orders).

Speedup: ~70% (wenn Orders selten gebraucht)


10. SQL Trace nutzen

Bottlenecks finden:

In ADT:

  1. RunRun Configurations
  2. Tab Trace Requests
  3. SQL Trace aktivieren
  4. Programm ausführen

Analysieren:

❌ SELECT * FROM mara
Duration: 2.5s
Rows: 500,000
✅ SELECT matnr FROM mara
WHERE mtart = 'FERT'
Duration: 0.01s
Rows: 1,000

Action: Langsame SELECTs optimieren!


Performance Checkliste

Code Review

  • Kein SELECT *
  • WHERE-Clause nutzt Index
  • Kein SELECT in LOOP
  • EML gebatcht
  • FOR ALL ENTRIES prüft auf leere Tabelle
  • CDS Views für Joins
  • FILTER/REDUCE statt LOOP
  • Buffering für Master-Daten

Measurement

  • SQL Trace durchgeführt
  • Performance < 1s für Standard-Ops
  • Keine Timeouts
  • Memory-Verbrauch akzeptabel

Performance-Benchmarks

Beispiel: 10.000 Datensätze

OperationVorherNachherSpeedup
SELECT *2.5s0.3s (nur Felder)88%
SELECT in Loop15.0s0.5s (FOR ALL ENTRIES)97%
LOOP + IF0.8s0.2s (FILTER)75%
ABAP Join3.0s0.4s (CDS)87%
EML Loop5.0s0.6s (Batch)88%

Gesamt: Von 26.3s auf 2.0s = 92% Speedup!


🎯 Zusammenfassung

Quick Win Ranking:

  1. ⭐⭐⭐ SELECT in Loop → FOR ALL ENTRIES (+90%)
  2. ⭐⭐⭐ EML Batching (+85%)
  3. ⭐⭐⭐ WHERE-Clause nutzen (+95%)
  4. ⭐⭐ CDS Views für Joins (+80%)
  5. ⭐⭐ Nur benötigte Felder (+60%)
  6. ⭐⭐ Lazy Loading (+70%)
  7. FILTER/REDUCE (+20-30%)
  8. Buffering (+95% aber niche)
  9. SQL Trace (findet Probleme)
  10. Projection Views (+60%)

Mindset: DB-Power nutzen, ABAP-Arbeit minimieren!


Siehe auch:

Happy Optimizing! 🚀