Ultimate Guide for Transformation Caches in Informatica IICS

---

Transformation caches in Informatica IICS are temporary storage areas that help Informatica Intelligent Cloud Services (IICS) process data faster. IICS first tries to store cache data in memory, and if memory is not enough, it writes the extra data to disk cache files. Understanding how these caches work will help you improve mapping performance and avoid failures.

1) What Is a Transformation Cache in IICS?

Transformation Caches in Informatica IICScache is a temporary storage area that IICS uses while running transformations. Its purpose is to speed up processing.

Two types of storage may be used:

✅ 1) Memory Cache (RAM)
Fastest.
IICS first tries to keep cache data here.

✅ 2) Disk Cache Files
Slower.
Stored under $PMCacheDir and $PMTempDir.

If memory is not enough, IICS writes overflow data to disk cache files, which slows the mapping.

Transformations that use caches:

Lookup, Joiner, Aggregator, Rank and Sorter

Except Sorter (which uses one combined cache), all others use two caches:

Index Cache → stores keys
Data Cache → stores values

---

2) How Transformation Cache Works (Diagram)

The flow is simple: IICS receives input data → runs a transformation → stores keys in Index Cache and row data in Data Cache → spills to disk if memory is insufficient → sends output rows to the next step.

---

3) Index Cache vs. Data Cache- Simple Examples

Below we will show for each transformation what goes Into Index Cache vs. Data Cache (With Clear Examples)

A) Lookup Transformation:

Lookup Table:

EmpID	EmpName	DeptID
101	Alice	10
102	Bob	20
103	Carol	10

Condition: EmpID = input.EmpID

• Index Cache: 101→R1, 102→R2, 103→R3
• Data Cache: R1=[Alice,10], R2=[Bob,20], R3=[Carol,10]

B) Joiner Transformation:

Master Table:

DeptID	DeptName
10	Sales
20	Finance

Detail Table:

EmpID	EmpName	DeptID
101	Alice	10
102	Bob	20

Condition: Master.DeptID = Detail.DeptID

• Index Cache: 10→R1, 20→R2
• Data Cache: A=[Sales], B=[Finance]

C) Aggregator Transformation

Input:

DeptID	Salary
10	1000
20	2000
10	1500

Condition: Group By DeptID
Expression: SUM(Salary)

• Index Cache: 10, 20
• Data Cache: 10→2500, 20→2000

D) Rank Transformation

Input:

EmpID	Salary
101	5000
102	8000
103	6000

Condition: Rank Top 2 salaries

The Data Cache stores sorted rows: (102,8000), (103,6000), (101,5000).

E) Sorter Transformation (Single Cache)

EmpID	Salary
101	5000
102	8000
103	6000

Condition: Sort Salary by DESC

(8000 → Row102)
(6000 → Row103)
(5000 → Row101)

---

4) Common Problems & Root‑Cause Analysis(RCA)

🔶Problem 1: High Precision Causes Slow Cache Creation

When a field has a very large precision (for example, a text field sized at 32,000 characters) as below image, IICS must allocate much more memory for the cache. This dramatically increases the time taken to build the cache and, in some cases, the task may appear stuck or even fail due to limited memory.

Even if you are not storing that many characters in the actual field, the Secure Agent still reserves space based on the precision definition. This means:

• Larger precision → larger cache structures
• Larger cache → slower build timePossibly,
• cache spills to disk → even slower performance

🔶 How to Improve Cache Efficiency (Action Steps)

✔ 1. Review precision of all incoming fields

• Look for text or large numeric fields with very high sizes.
• If their actual data is much smaller, reduce the precision.

✔ 2. Check return fields in Lookup transformations

• Avoid returning very wide fields if not needed.
• Return only what is necessary.

✔ 3. Inspect Sort Keys in Sorter

• Only sort on columns required for your logic.
• Avoid sorting on wide string columns unless required.

✔ 4. Check Master Pipeline of Joiner

• Remove oversized fields before they reach the Joiner.
• Use Expression transformations to shrink precision early in the flow.

🔶Problem 2: Cache Folder Gets Full in IICS

Whenever you click Preview inside a mapping, IICS quietly generates temporary CSV files behind the scenes. These files are stored in the Secure Agent under:

.../apps/Data_Integration_Server/data/cache/preview

Here’s what actually gets written:

• One CSV file for the transformation you previewed
• One CSV file for each upstream transformation
• One file per output group, if the transformation has multiple outputs

If you preview many times during mapping development, this folder can quickly grow in size.

But IICS automatically cleans these preview files every 24 hours, removing anything older than a day. However, if you generate a lot of previews within a short time window, the folder may temporarily appear full until the automated cleanup happens.

🔶Problem 3: Cache Files Left Behind After Failed Mapping Runs

During actual execution, transformations like Joiner and Aggregator generate cache files with specific prefixes:

• PMAGG* → Aggregator cache files
• PMJNR* → Joiner cache files
• PMLKP* → Lookup cache files

Normally, IICS deletes these automatically after a successful run. But if the mapping fails midway, the cleanup process does not occur. This means:

• Stale cache files remain in the directory
• The folder gradually fills up
• Disk space warnings or failures may occur later

---

5) How to Find Correct Cache Size Using Logs

Step 1 — Set tracing to Verbose Initialization

Transformation → Advanced tab →
Tracing Level = Verbose Initialization

Step 2 — Run mapping in Auto Cache size mode

Step 3 — Look at session log
You will see entries like below:

These are the exact sizes your transformation needs.

Step 4 — Set these values manually

Now you enter them in the transformation’s Advanced tab, so the cache fits in memory. This prevents disk spilling.

---

6) Tuning Tips

• Lookup → Reduce precision, filter SQL
• Joiner → Smaller Master, sort inputs
• Aggregator → Fewer group‑by columns
• Sorter → Sort only needed columns
• Disk cache → Clean stale files regularly

---

7) Summary Table

Transformation	Index Cache	Data Cache
Lookup	Lookup keys	Lookup rows
Joiner	Master join keys	Master rows
Aggregator	Group keys	Aggregate values
Rank	Group keys	Sorted rows
Sorter	—	Sort keys + rows

---