0

I need to get an understanding of ISO-8583 message platform,lets say i want to perform a authorization of a card transaction,so in real time at a particular instance lets say i got 100000 requests from network(VISA/MASTERCARD) all for authorization,how do i define priority of there request and the response,can the connection pool handle it(in my case its HIKARI),how is it done banks/financial institutions for authorizing a request.Please provide me some insights on how to manage all these requests.Should i go for a MQ?

Tech used are:-spring boot,hibernate,spring-tcp-starter

2 Answers2

0

Your question doesn't seem to be very well researched as there are a ton of switch platforms out there that due this today and many of their technology guides can be found on the web including for major vendors like ACI, FIS, AJB,.. etc if you look yard enough.

I have worked with several iso-interface specifications, commercial switches, and home grown platforms and it is actually pretty consistent in how they do the core realtime processing.

This information on prioritization is generally in each ISO-8583 message processing specification and is made explicitly clear in almost every specification I've ever read written by someone who is familar with ISO-8533 and not just making up their own variant or copying someone elses.

That said.. in general at a high level authorizations / financials (0100, 0200) requests always have high priority than force posts (0x20) messages.

Administrative messages in the 05xx and 06xx and 08xx sometimes also get bumped up above other advices.. but these are still advices and almost always auths/financials are always processed first as they A) Impact the customer B) have much tighter timers than any advice by usually more than double or more.

Most switches I have seen do it entirely in memory without going to MQ and or some other disk for core authorization process to manage these.. but not to say there is not some sort of home grown middle ware sometimes involved.. but non-realtime processes regularly use a MQ process to queue or disk queuing these up into processes not in-line of the approval for this Store-and-forward (SAF) processing.. but many of these still use memory only processing for the front of their queue.

It is important to also differentiate between 100000 requests and 100000 transactions.. the various exchanges both internal and external make a big difference in the number of actual requests/responses in flight at even given time.. a basic transaction can be accomplished in like two messages.. but some of the more complex ones can easily exceed 20 messages just for a pre-authorization or a completion component.

If you are dealing with largely batch transaction bursts.. I can see the challenge of queuing but almost every application I have seen has a max in flight for advices and requests separate of each other.. and sometimes even with different timers.. and the apps pumping the transactions almost always wait for the response back before sending more.. and this tends to work fine for just about everyone.. including big posting batches from retailers and card networks. So if your app doesn't have them.. you probably need to add them.

CRSouser
  • 658
  • 9
  • 25
0

In fact your 100000 requests should be sorted by (Terminal ID and/or Merchant ID) + (timestamp/local timestamp) + (STAN and/or RRN). Duplicated transaction requests expected to be rejected.

If you simulating multiple requests from single terminal (or host) with same test card details the increasing of STAN/RRN would be a case.

Please refer to previous answers about STAN and RRN ISO 8583 fields. In ISO message, what's the use of stan and rrn ?

iso8583.info support
  • 2,130
  • 14
  • 18