Friday, October 17, 2008

What could be wrong with the new Maybank2U.com?


Looks like the team at MBB decided to switch on back their previous version of the site for temporary interim solution while they "fixing" the performance issues. Man, the old one looks so efficient now compared to the new boy.

Well, I read somewhere about the cost of the new site is around MYR5 Million.

Does the new Maybank2U.com value added to the bank? Frankly speaking, I'm not sure because I can't tell what else differentiating the versions besides the intended Web 2.0 facelifts which turned out to be a "plastic surgery" disaster at the moment.

Does the new one supports more transactions per user, per second, per day or per account?

Does it promises a lower total costs of ownership (TCO) over next 5 years?

Can it scale better to support more concurrent transactions or users?

Are users happier? Better user experiences?

Is the Maybank brand stronger?

....

From a pure business point of view, I really need to quantify the costs and benefits of this. Would you pay 40k extra for a new model of Toyota Vios which merely a facelift from last year model? Not me.

The worst case scenario from unnecessary spendings/expenditures is the consumers need to bear the consequences, in other words, the cost is transferred to us!!! Do we still remember the day when MBB almost imposes A RM10 i-forgot-what-kind-of-fee for the usages of Maybank2U.com. Given the current circumstances, I believe the probability of imposing such non-interest fee to us is changed from "likely" to "highly likely".

Sad. 5 Million for this and another 8+ Billion for BII....

Another way, let's jump out from the mud of business conspiracies and make some educated guesses about what could be wrong with Maybank2U.com technically.


1. Hardware Sizing issues
They might have made unrealistic sizing benchmarks using previous version performance data. This might due to underestimation of overhead of selected platforms and methods. This problem is easiest to solve: Plug in more juices would do. More RAM, More CPU and More HDD/Disk Controller, even more cluster members. Of course, care must be taken to identify the bottleneck currently.

2. Networking issues
Related to sizing. Unexpected loads on routers, backbones, swicthes might causes performance trashes. It could also be configuration issues on the networking components that results in erroreous or inundated communication overheads between members in the network.

3. Messaging issues
Not sure what kind of messaging facilities are in used in the environment but the new system might be making some messaging assumptions that are not optimized for the existing messaging infrastructures.

4. Software issues
Web server, application servers, containers and others might need to be further tuned. Design limitations of certain servers/components of the server might be problematic when the system requires more than permitted. For example, some version of application server may be have a 2GB JVM memory limit which could causes performance trouble when the application is badly designed that it requested more than the limit most of the time. Thread pools, Process pools, Block sizes and many other factors may within consideration.

5. Framework issues
SOAP listener ports or any AJAX request receivers must be properly tuned. Apache Struts framework performance enhancing features should be considered. Seggregation of static components and dynamic components of the application in clustered web servers and application servers is recommended. Server Caching facilities can be utilized. Logging facilities and other development stuffs should be turned off to suit production environment.

6. Human issues
Sabotages? Less likely.

Performance issues and troubleshooting after roll-out is usually difficult in the sense that when the team turned on the benchmarking, log tracing, monitoring and diagnostic tools, you can expect a more drastic performance degradation until the point that the system is virtually usable from a user point of view and yet you might be crossing your fingers hoping the true problems are identified.

It is so entertaining to see how a Web 2.0 application (Expensive one too) undresses its pretty covers and shows cast the obstreperous of Web -1.0 and ignites the wrath of peasants.

Good luck and make sure you doesn't miss out a ZERO in my checking account balance.