MegaCardStandalone

Online credit card processing with integrated fraud checking as part of a Benchmark. Written in Java

It requires a SQL Database (DB2) with the schema defined in schema.ddl

Requests

The application has one HTTP resource to accept incomming requests:

POST /MegaCardStandalone/MegaCard/Svc/Transfer

The request body is json formatted:

{
    "merchantAcc": ${merchantAcc},
    "merchantToken": "${merchantToken}",
    "cardNumber": "${cardNumber}",
    "cvv": "${cvv}",
    "expirationDate": "${expirationDate}",
    "amount": ${amount},
    "transactionUuid": "${__UUID}",
    "method": "${method}",
    "timestamp": ${timestamp}
}

timestamp is optional and only used to make runs reproducable

Response

If the transaction was successful the string successfulis returned.

Cofiguration

The app can be configured through a set of environment variables:

Fraud Checking Basic

Variable	Description
CHECK_FRAUD	(true/false) Enable / Disable Fraud checking and quering the required history from the Database
ModelAdapterClass	Class which implements the FraudChecking. Needs to implement the com.ibm.lozperf.mb.ModelAdapter interface. The App comes with a number of Adapters in com.ibm.lozperf.mb.modeladapter: TFServingAdapter, TFServingBatchingAdapter, DLCModelBatchingMTPPAdapter

Tensorflow Serving

Varaible	Description
TF_URL	Url of the tensorflow model server to be used by TFServingAdapter

Batching

Model Adapters with Batching in the name use com.ibm.lozperf.mb.batching.BatchCollector to implement batching of fraud checking requests. The BatchCollector can be configured through the following Variables

Variable	Description
PREDICT_THREADS	Number of Threads to concurrently process the batches
TARGET_BS	Size of the batches the BatchCollector tries to archive. If no Predict thread is availible, the batches can grow brigger.
BATCH_TIMEOUT	Maximum time to wait for additinal requests, after the first request received before processing of the batch is triggered as if TARGET_BS is reached.

DLC Model Adapters

The DLC doesn't implement string operations at the moment, so the strings first need to be mapped to an Integer Index. These mappings are read from CSV files using the |as the seperator. For example:

[UNK]|0
Money Transfer Company 1|1
Convenience Store Chain 1|2
Auto Toll Provider 1|3
Supermarket Chain 2|4
Wholesale Club 1|5
Convenience Store Chain 3|6
Gas Station Company 1|7
Supermarket Chain 3|8
Bookstore Company 1|9

The name of the mapping files are: MCC.csv MerchantCity.csv MerchantName.csv MerchantState.csv Zip.csv

The directory of the mapping files is specified through the STRING_MAP_DIR variable.

JMeter

MegaCard_date.jmxcan be used to drive requests against the application.