text bug da aprire ad orika

Posted 2021-04-30

A stress test was performed on the ascent of the receipts, with the objective to understand how many receipts can be recovered at the same time.

The environment was formed in this way:
2 orkcentral application with 4GB RAM
10 orkaisse application with 4GB RAM
5 istances of orkaisse5 on every orkaisse application to simulate 50 shops

PREREQUISITE
To perform the test we had to comment the call of the writeReceiptToDB function in the on_BROKER_RECEIPT_RECEIVED. Because otherwise only 10/20 tickets per shops ascend.

TEST 1
Number of shops: 50
Number of tickets per shop: 1000
Delay time between ascent: 0.250s
Max DB connection: 100
Number of DB connection used on orkcentral: 22

Result of the test: orkaisse5 fill the RAM and the swap memory of both the orkcentral application and they both crash. Ticket received by orkcentral: 4000

-------------------------------------------------

TEST 2
Number of shops: 10
Number of ticket per shop: 5000
Tickets received by orkcentral: 30500
Delay time between ascent: 0.250s
Max DB connection: 300
Number of DB connection used on orkcentral: 22

Result of the test: crash of the orkcentral application 2. Instead the orkcentral application 1 after a few minutes of stop succeeds in integrating all the tickets.

-------------------------------------------------

Increasing the delay time between the ascent to 0.500 improved the performance, but do not solve the memory problem. Consider also that orkaisse5 isn't the only service that can require a lot of memory running on Orkcentral.