Productive suggestions for 10C at Peak Traffic

benjs · April 2015

Rather than adding to the complaints list, I figured I'd try to offer some ways 10C could fix some of their problems.

This is a problem that many websites have: their peak traffic represents a massive multiple of what normal requirements are. While one or two servers might be sufficient for day-to-day web hosting, days like today represent times where 7 or 8 servers might be necessary to adequately handle bandwidth requirements.

Three suggestions off-hand:
1) Amazon Web Services. See this white paper to better understand "on-demand provisioning of additional servers, which allows capacity and costs to follow the traffic model". In other words: pay for your traffic. If the traffic comes, your serving abilities follow, and you never let your customers down. This is simple, utilizes money appropriately, though might have a greater per-bandwidth cost. That being said, the savings are seen in customer dissatisfaction - which, if it isn't measured yet by Ten Club, seriously ought to be. See https://s3.amazonaws.com/aws001/trailhead/AWS_Web_Hosting_Best_Practices.pdf for more details on how AWS can assist Ten Club.

2) More serving power. This is a good idea, as it accommodates the customer's needs. On the other hand, it is a poor economic decision as it wastes money on additional servers which are only necessary for peak traffic opportunities.

3) Deli-style. If you walk into a popular deli, you take a ticket, and wait until your number is called. This is based on your position in line, which is based on when you showed up. If two (or three, or in the case of Pearl Jam fans - ten thousand) people show up at exactly the same time and try to walk through the door at once to take a ticket, they create a bottleneck and no one gets through: not them, not the people behind them. At this point, we need the meat counter man (aka Ten Club employee, but better for it to be a randomizing application) to count the people stuck in the doorway, give them each a ticket one by one randomly, usher them into the store, then continue to allow people in based on the order they arrived (with more chronologically assigned tickets). When the store is full (or the meat/mystery vinyl quota is filled), the people in line should be notified that if they don't run out of meat (or mystery vinyl packages), their place in line will be maintained (if they wish to wait), and that they will be notified when that happens.

Being honest here - Ten Club runs a very tiny deli. What it might be necessary to do is to look at the doorway that can accommodate one person at a time, and start to use the same logic on the lineup going down the block: assign tickets, and explain to people that the shop is low on space, so they're going to have to wait outside in the store's 'gateway (a very low-bandwidth required text-only page for those in the queue)' until the space (bandwidth) is adequate to allow entry.

There is a fourth option, and I bring it up so that there's no ambiguity here.

4) Do nothing. In this option, the band and organization states in their silence that they don't measure customer happiness or dissatisfaction (aka they see dollars coming in, and that's what matters). The things that a business doesn't measure are the things that a business doesn't care about. In this case, customer happiness or dissatisfaction would make that list of "frankly, we don't care". I hope options 1-3 (or another I haven't thought of) are seriously considered.

Curious to hear others' thoughts!