Office Booker

The Office Booker was created to solve the problem of coordinating the safe return to offices once COVID-19 restrictions are lifted.

The Lab team of 4 developers put together this real-time booking app in just 10 days for proactively managing demand for office space by allowing you to book ahead of travelling to the office. Although originally created as a tactical solution within Telefónica UK but we decided to share this publicly as it might be a problem other companies are also facing - including some of our own customers.

Contributors

Getting Started

Prerequisites

Node.js, Yarn, Pulumi, Docker, AWS CLI

Local Quick Start

1. ./install.sh - Restore packages
2. ./start.sh start API and client application

The default local auth setup allows you to enter any email address and code.

Note: Run ./make-env.sh [STACK] to use config from the deployed stack for local testing.

User Journey

1. Log in with email
2. Pick office and day - showing which are available or not
3. Click to reserve a space
4. Can cancel reservation if needed
5. Virtual pass shown in app or printable which security visually verifies

System Rules

• Bookings are for the whole day
• Users can only make 1 booking per office, per day
• Default limit of 1 booking per week per user, can be adjusted by System Administrators
• Bookings for the today can only be cancelled by Administrators
• Users pick a single office (can be changed in help)

First Deploy

Before deploy

1. Register domain
2. Add Zone to Route53
3. Change NS records to created Zone

Deploy

1. ./install.sh - restore all package dependencies
2. ./build.sh - audit, test and build infrastructure assets
3. cd infrastructure - this is where deployment is coordinated
4. pulumi login - this is where the state of your deployment is stored
5. pulumi stack init - create a new stack
6. Configure stack settings (see Pulumi Config Example below)
7. cd .. - Move back to the root of the project
8. ./deploy.sh [STACK] - deploy the stack created in step 5

After deploy

1. Add verified address
2. Open Cognito Pools -> office-booker-booking-users-xxxxx -> FROM email address -> edit
   • Do you want to send emails through your Amazon SES Configuration? - Yes - Use Amazon SES
   • Fill required files

Note: the first smoke test might fail as DNS entries can take a while to propagate.

Subsequent Deploys

This is the same as for the first deploy, excluding creating and configuring the stack:

1. ./install.sh
2. ./build.sh
3. ./deploy.sh [STACK]

Privacy Policy

Before running this application, you should review the privacy policy and adjust for your own situation.

Releases

We are following semantic versioning and maintain a changelog to indicate the impact of changes.

Each major release must be installed in order to make sure that migrations are completed. I.e. if you're running version 3 and want version 5, you must deploy version 4 first.

Pulumi Config Example

config:
  aws:region: eu-west-1
  office-booker:advance-booking-days: '14'
  office-booker:default-weekly-quota: '1'
  office-booker:dns-zone: my-company.example.
  office-booker:domain-name: my-office-booker.my-company.example
  office-booker:email-regex: ^(.*)@(my-company\.example)$
  office-booker:office-quotas:
    - name: Office Alpha
      id: alpha
      quota: 100
      parkingQuota: 60
    - name: Office Beta
      id: beta
      quota: 200
      parkingQuota: 150
  office-booker:registration-from-address: [email protected]
  office-booker:selftest-key:
    secure: v1:xxxxxxxxxxxxxxxx:xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx==
  office-booker:selftest-user: [email protected]
  office-booker:system-admin-emails:
    - [email protected]
    - [email protected]
  office-booker:reason-to-book-req: true
  office-booker:notification-to-address: [email protected]

Note:

The secure selftest-key can be generated and set by running:

pulumi config set --secret selftest-key `openssl rand -base64 30`

If setting reason-to-book-req as true, you must also supply a notification-to-address.

User Roles

Quotas are applied to all users regardless of role.

• Default: Any user with a valid email address gets this role.
  • Can manage their own bookings only.
• System Admin: Must be configured in Pulumi
  • Can manage all bookings in the system
  • Can manage other users
• Office Admin: Must be assigned by a System Admin
  • Can manage bookings for their assigned offices
  • Can view other users (but can't make changes)

Logging

General meanings of log levels:

• ERROR: something broke in the server which should have worked fine
• INFO: everything worked as expected, but the user might have been shown an error

All changes to the database are audit-logged for trouble-shooting purposes.

Solution Design

Technical Summary

• Email OTAC login, limited by configured regular expression
• Online administration dashboard for booking and user management
• Designed for use on mobiles but has desktop support too
• Linear cost scaling based on usage
• All storage is encrypted-at-rest
• No remote terminal access or system patching required

AWS Cloud Infrastructure

The client application is a React.js single page application (SPA). We uses AWS Cognito for authentication. The API is written on NodeJS running on AWS Lambda via API Gateway with dynamic data stored in DynamoDB (encrypted via AWS KMS).

Monitoring

All logged errors are centrally monitored by the alerts lambda. Add your own code to alerts/src/lambda.ts in the postNotification function to send notifications to somewhere like a slack channel or email account.

Service Shutdown

1. Disable the CloudFront CDN to stop all traffic
2. Empty the S3 api-cdn-logs and api-static-site buckets
3. Run pulumi destroy

Cost Estimation

Due to the serverless architecture of the application, all costs scale linearly based on:

1. The number of monthly active users
2. The number of requests to the application

If no requests are made to the system, it should cost $0.00 USD per month.

Here's a detailed example of estimated cost based on the following generous approximations:

• 10,000 users
• 1,000 API requests per user, per day (extremely heavy usage)
• 100 milliseconds to process each request
• 1KB API response size
• London AWS region

Upper-bound estimated cost per month, excluding free tier: $494.38 USD

Breakdown by service:

• Lambda: $310.00
  • Requests (300,000,000): $60.00
  • Compute Time (30,000,000 seconds): $250.00
• S3: $0.07
  • Standard Storage (1GB): $0.02
  • GET Requests (100,000): $0.04
• Cognito (10,000 Monthly Active Users): $55.00
• API Gateway (300,000,000 API Calls): $105.00
• DynamoDB: $0.00
• CloudFront (286 GB out): $24.31

Estimated cost, including free tier is $428.51 USD:

• Lambda:
  • 1,000,000 free requests per month
  • 400,000 GB-seconds of compute time per month
• Cognito: First 50,000 users are always free
• S3 - free for first year:
  • 5 GB of Standard Storage
  • 20,000 Get Requests
  • 2,000 Put Requests
• API Gateway: 1,000,000 API Calls Received per month for first year
• CloudFront: 50 GB out free for first year