Distributed Vs Centralized Systems in the Mastercard Organization

The MasterCard (MC) Global Technology and Operations facility located in St. Louis, Missouri is at the center of the company's operations. Since its upgrade in 19971, MasterCard has invested more than $136 million dollars in that facility, which consists of 743 miles of conduit, 412 miles of wire, 100 miles of copper cable, 160 miles of optic fiber, 2.8 miles of cable tray underneath raised floors, 360 data cabinets and 652+ terabytes of storage2. This facility is a major expense for the MasterCard organization, although it is not their primary source of revenue. Keeping in mind that the main revenue source is rather the data mining and processing software that the company provides, that central processing facility becomes a costly overhead with the new equipment upgrades that are currently required.

A two part suggestion has been made for a restructuring reform of that main facility. One: the implementation of a distributed database system, in which the member banks will hold part or a whole copy of the database and two: the use of commercial of the shelf (COTS) software developed by MC and distributed to the members for local use, rather then a centralized processing application server.

Each of those two propositions, taken separately or together, brings to light a number of benefits and drawbacks compared to the current model. In order for one to make a justified decision one needs to familiarize himself with those ramifications.

DISTRIBUTED DATA MODEL

The distributed data model has a lot of advantages over a large centralized system that MC currently uses. This model eliminates the needs and expense for central storage. The cost associated with MasterCard's central storage facility will be obsolete therefore reduction in monthly fees for service can propagate to member institutions and individual cardholders. If the data is partitioned among member banks the local storage will increase response times. The size of the local portion of the database will be determined by the individual member bank needs.

The expansion of the network is easy as new members obtain a local copy only of the portion they need.

Another advantage of this distributed model is that it reduces traffic overhead for as long as the majority of queries to the database are local to the copy that a specific member bank has on hand. This model creates autonomy. If one member goes off line for any kind of technical or maintenance reasons the rest can continue to function independently.

Unfortunately together with the shorter response times and greater expandability there are several unavoidable drawbacks to the distributed data model. Data consistency is going to be a problem since each member bank has the ability to modify their local copies without everybody else finding out about the change in a timely fashion. Although there are schemas and known algorithms in place that can accommodate for real-time updates and propagation the communication overhead will be very costly and critical. Very elaborate effort is required to maintain data up-to-date, and may require such highly specialized staff that the cost may not make economical sense for neither the member banks nor the MasterCard organization itself.

Response time will suffer dramatically if daily queries require access to multiple DB instances stored at different geographical locations. Time associated with finding where the appropriate data is stored, time authenticating to the appropriate data server, and time retrieving that data may add up to a point where it becomes taxing for the business process.

Security is another major issue as with any distributed system. Since each member bank server needs to have access to every other member bank's DB server a very complicated authentication algorithms are needed that can create both traffic overhead, slow response times and potential security vulnerabilities. The problem with this kind of a model is that the whole system is only as strong as its weakest link. One compromised member could propagate security issues to all members. Many cardholders feel more comfortable with their data being stored with large and powerful organization such as MasterCard compared to a local bank that may not necessary have the IT budget to secure their data as well as a larger bank or MasterCard could.

A security compromise as the one described above will have multiple implications. A smaller bank member may not be able take such a "hit" if all cardholders take legal action against it. Small banks will start going out of business which is bad for them and for MC. Large banks may benefit from this condition.

Another implication is the public opinion and MasterCard's over all market share. If a small local bank gets compromised from a security stand point the news will get out that MasterCard holders from that bank have a serious problem. Although it was cardholders from that specific bank that were affected the MC logo gets associated with negative PR and in the long run the general population will slowly but surely start to lose confidence in MasterCard and in result hurt MasterCard's market share. Although operational cost will be reduced for MC this may not be a successful business plan in the long run.

With the distributed data approach MasterCard comes out as an immediate winner. Huge maintenance cost in both equipment and personal can be eliminated and the IT team can be restructured so that in focuses on MasterCard's core functionality of providing data mining tools VS spending a great deal of time and resources in DB management and upkeep.

Member banks are the immediate "losers" in this schema. They will be required to purchase and maintain equipment that was not needed before, and furthermore they will not have total control over their own equipment since the need for standardization will be met by following strict specifications mandated by MasterCard.

A positive of this restructuring approach is that the end users (i.e. member banks personal) will not be affected regardless of such a major organizational change. Applications will continue to function in the same manner that they did before the change. This will at least theoretically not affect the business process in any significant or negative way. Response times and traffic overhead can dramatically improve as long as the majority of queries are local to that member bank. Unfortunately the data integrity will most likely suffer.

If a system is distributed finding where the problem occurred could be just as complicated task as fixing that very problem. The end user would most likely have a negative reaction to the fact that their credit identity is stored in a potentially less secure institution