Skip to main content

Case Study: T-Mobile Tuesdays and The Starbucks Deal of December 10th, 2019

SITUATION: About 3 weeks ago, T-Mobile has a second deal in its popular T-Mobile Tuesdays app:

"Come back at 3PM Central to get a Starbucks 3$ gift card!"

This would not be the first time the Telecommunications brand encouraged people to flood the app to get free or discounted food - earlier in 2019, there was a similar deal with Panda Express, and the Chinese-American Chain was not prepared to handle the orders, letting many of them go unfulfilled.

Would T-Mobile be more ready? I mean, they are a telecommunications company, right?





I won't pretend that I know how deeply different computer networking infrastructure and telecom infrastructure are. But I do recall seeing a tweet in the 30 minutes or so of refreshing I (and many others) were doing -

"Instead of teaming up with Sprint, T-Mobile should have teamed up with Azure or AWS."

Referring to flexible cloud computing. So I thought, well, here's how I could fix that.

For the unitiated, T-Mobile Tuesdays is a weekly 'freebie' giveaway on the aforementioned app on, you guessed it, Tuesdays.

It's often something small, like a free burrito from Taco Bell, or a free rental from Redbox. My favorites are the variations on movie ticket discounts from Atom Tickets - They vary from 5$ off with no fees to (the rarer) 5$ for ANY movie ticket period. I used the latter to get a 17$ ticket for 5$.

While the goodies come from other retailers, T-Mobile pays for them upfront, and provides the user with a code to apply on the merchant's site through the app.

So, why did the Starbucks giveaway go so badly when, most of the time, the app works perfectly fine?

Two words: Limited Quantities.

Most deals on the app simply end at the end of the day. With how popular the glorified milkshake bar is, many people were excited to have $3 off, and caused services to overwhelm with demand.


That sounds like something a scale set should be able to handle. The number of VMs will automatically increase or decrease in response to demand. Putting the VMs in the same Virtual Network means if one happens to go down (somehow), load balancing can maneuver to the others.



Here are the basic stats.



There's only 10* for now. In a real T-Mobile production environment, I'd have to see the average throughput of hits / data crunched an hour and work from there.

There's an option to enable scaling beyond 100 instances if need be, but at most, we can use 20. I can also enable autoscaling based upon CPU usage.

Allow select inbound ports to be open. After all, we need people to interact with the server to get the codes. I do allow some more inbound security rules:


Load balancers will need to be attached to Virtual Machines and Virtual Networks.



In very (very) brief, there is one possible solution in scaling out resources to make sure there's enough for people to access.



* I kept getting hit with Azure limits 'to avoid misuse', so later screenshots will definitely have less than 10/20
EMPLOYERS: It's a game plan on what I can do.

Comments

Popular posts from this blog

Connecting IoT Devices to a Registration Server (Packet Tracer, Cisco)

In Packet Tracer, a demo software made by Cisco Systems. It certainly has changed a lot since 2016. It's almost an Olympic feat to even get started with it now, but it does look snazzy. This is for the new CCNA, that integrates, among other things, IoT and Automation, which I've worked on here before. Instructions here . I don't know if this is an aspect of "Let's make sure people are paying attention and not simply following blindly", or an oversight - The instructions indicate a Meraki Server, when a regular one is the working option here. I have to enable the IoT service on this server. Also, we assign the server an IPv4 address from a DHCP pool instead of giving it a static one. For something that handles our IoT business, perhaps that's safer; Getting a new IPv4 address every week or so is a minimal step against an intruder, but it is a step. There are no devices associated with this new server; In an earlier lab (not shown), I attached them to 'H

Securing Terraform and You Part 1 -- rego, Tfsec, and Terrascan

9/20: The open source version of Terraform is now  OpenTofu     Sometimes, I write articles even when things don't work. It's about showing a learning process.  Using IaC means consistency, and one thing you don't want to do is have 5 open S3 buckets on AWS that anyone on the internet can reach.  That's where tools such as Terrascan and Tfsec come in, where we can make our own policies and rules to be checked against our code before we init.  As this was contract work, I can't show you the exact code used, but I can tell you that this blog post by Cesar Rodriguez of Cloud Security Musings was quite helpful, as well as this one by Chris Ayers . The issue is using Rego; I found a cool VS Code Extension; Terrascan Rego Editor , as well as several courses on Styra Academy; Policy Authoring and Policy Essentials . The big issue was figuring out how to tell Terrascan to follow a certain policy; I made it, put it in a directory, and ran the program while in that directory

Building, Breaking, and Building A CRM with Retool

 I like no- or low-code solutions to things. I've often wanted to simply push a button or move some GUI around and have the code implement itself.  I've thought about building something that's like a customer relationship management (CRM) system for keeping up with my network better than my little spreadsheet where I click links and then go like something. The general idea in this CRM Development is:  To have a GUI to add people to a NRM (Network Relationship Management).       Attach it to a database (MySQL is what I went with eventually using Amazon Relational Database service, but you can use PostGRES, and probably others).     Make sure components are connected to each other in the retool interface. This video is a good start. Watching the tutorial video, heard some SQL commands and went 'Oh no 😳" before going "Wait I know basic SQL", which is good, because you'll see.  When you get set up, there's a plethora of resources you can use -- Incl