Hacker News Reader

Comment by iamd3vil | Hacker News Reader

Since I am an Indian I know of a service which is hated by most people - IRCTC because of it's scaling issues. It's a lot better now but I remember those horrible days you have to do a tatkal ticket.

I also know of some government services (my state has paperless administration so everything is digital) which are used in day to day basis (but only hundreds of users) by government employees goes down for almost 2 hours per day everyday which stops work for everyone (both the employees and the people came for that service). These services are developed by companies like TCS and Infosys. Only if they thought of scaling before.

View on HN

In case of government, it's often another issue. While the social network of the parent was probably a lean stack, the government one was probably over-engineered. Some webapps are front-ends that communicate with a back-end in XML, with massive flaws like serialization and lack of transactions, and the backend syncs with a business solution backed by dBase, all of that for 11 web pages (I know, I've been on govt apps before – France). Scaling was a major requirement from the beginning, but those engineers just don't know what they're doing and follow the XML frontend/ESB bus patterns, assuming that's what gives performance. And committee design (12 to 24 ppl) does the rest of the bloat. 11 screens, $2m, sluggish result.

Had they started with only $300 like the parent comment, their webapp would have often performed much better. Big projects are a difficult curse, our most difficult question in IT engineering.

> These services are developed by companies like TCS and Infosys

This is the real problem: lack of in-house expertise and ongoing incentives to maintain performance, while contractors are usually paid for features and, if they also do hosting, even have a financial disincentive for efficiency.

I would bet it's at least as likely that had someone thought of scaling it would have added a significant cost and delay to the actual project and the first major scalability bottleneck would still have been something unanticipated.

Reply on news.ycombinator.com