System requirements

Functional:

List functional requirements for the system (Ask the chat bot for hints if stuck.)...

• search for hotels and rooms based on location, dates and number of guests
• view a given room type
• make a reservation
• view a reservation
• cancel a reservation

Non-Functional:

List non-functional requirements for the system...

• highly available
• consistent
• high performance (several users can view and/or book simultaneously)
• responsive and interactive (~50ms page load time on the UI)
• Secure

Capacity estimation

Estimate the scale of the system you are going to design...

Max number of users browsing or viewing rooms per second = 10K

Max number of users viewing a room per second = 5K

Max number of users trying to book a room per second = 1K

Reads are much more than writes so this is a read-heavy system.

Total number of hotels in the system = 1K

Total number of rooms maintained in the system = 50K

Given that the data and number of concurrent users is not very high, an SQL Database like MySQL or Postgres will be suitable. SQL DB also provides transaction support and ACID properties which meet our requirements.

API design

Define what APIs are expected from the system...

• Get available room types based on search criteria

GET https:///rooms?start=x&&end=y&&numGuests=n

Returns array of JSON responses

• View a particular room

GET https:///room?type=t

• Make a reservation

POST https:///reserve?roomType=t&&start=s&&end=e

Response type: 201 (success) with reservation id or

40x(error) or 50x(error)

• Cancel a reservation

DELETE https://reservation?id=x

Response type 201 (success) or 40x(error) or 50x(error)

Database design

Defining the system data model early on will clarify how data will flow among different components of the system. Also you could draw an ER diagram using the diagramming tool to enhance your design...

Hotels

======

HotelId

Name

Location

Address

Rating

RoomType

=========

Type -- Primary Key

HotelId -- Foreign Key

Name

NumOccupants

TotalRooms

FeatureA

FeatureB

...

Rate

======

Date --- Primary Key

RoomType --- Primary Key

HotelId

Rate

Rooms

========

RoomId --- Primary Key

HotelId

Type

Status (InMarket/OffMarket)

Number

Location

RoomTypeAvailability (used for conflict detection in concurrent booking)

======================================================

RoomType --- Primary Key

Date --- Primary Key

NumAvailable (unreserved) --- Primary Key

HotelId --- Foreign Key

Reservation

==========

ReservationId ---- Primary Key

Name

RoomType

HotelId. ----- Foreign Key

NumGuests

Start Date

End Date

Status (complete/pending/canceled)

Payment

========

PaymentId. ---- Primary Key

ReservationId --- Foreign Key

Amount

Status

Notes

High-level design

You should identify enough components that are needed to solve the actual problem from end to end. Also remember to draw a block diagram using the diagramming tool to augment your design. If you are unfamiliar with the tool, you can simply describe your design to the chat bot and ask it to generate a starter diagram for you to modify...

External User: End-users which use the booking system for viewing rooms, booking, cancellation, feedback etc.

Admin User: Internal Hotel Mgmt Staff has access to certain APIs for adding/removing a hotel, adding/removing rooms, rates of rooms etc.

Hotel Data: Aggregator which accepts rooms and other data from hotels. The booking system has several hotels as external partners and it can show/book rooms for these hotels.

User Mgmt: User sign up, sign-in, login etc.

Rate: Rates of rooms are decided by the Rate Service. The rates can be changed dynamically based on peak season, location etc.

Rooms: Management of rooms and Rooms table

Reservation: Room reservation, booking and cancellation.

Payment: Accepts payments for booking. It interfaces with an external Payment Service Provider(PSP) which does the actual transaction.

Feedback/Rating: Takes user feedback or rating for a hotel, room etc.

Recommendation: Recommends rooms, hotels and destinations based on past bookings and popular reviews. Can use Clustering Algorithms or Collaborative Algorithms for recommendation.

Cache: Frequently used and static data can be in a cache implemented in Redis or a similar technology. For example, Rate information for the next 2 months can be in the cache.

Notification Service: Used to notify users of rate changes for a room, upcoming and successful reservations etc. Also notifies individual hotels about new bookings and cancellations.

These services are stateless and can be deployed as microservices to support higher scale.

Request flows

Explain how the request flows from end to end in your high level design. Also you could draw a sequence diagram using the diagramming tool to enhance your explanation...

User searches for rooms based on location, dates and num guests

=================================================

The booking system performs the following DB queries:

• types = select roomtype from RoomType where numGuests >= n;
• availableTypes = select roomtype from RoomTypeAvailability where type in {types} && date>=start && date <= end && numAvailable > 0
• select roomtype, hotelId, rate from Rate where roomtype in {availableTypes}

View a given room type

==================

• select * from RoomType where roomtype = type && hotelId = h

Make a reservation

================

Concurrency in booking the same room should not lead to data inconsistency. To avoid this, the RoomTypeAvailability table is used.

• When a booking is done, this table is updated as follows for each date:

update RoomTypeAvailability set numAvailable = numAvailable-1 where roomType = t && date = d

• As numAvailable is a primary key, concurrent updates to the table will not be allowed.
• Atomically, the Reservation table is also updated in the same transaction.

Detailed component design

Dig deeper into 2-3 components and explain in detail how they work. For example, how well does each component scale? Any relevant algorithm or data structure you like to use for a component? Also you could draw a diagram using the diagramming tool to enhance your design...

• To avoid inconsistency in concurrent booking, the RoomTypeAvailability table is used.

When a booking is done, this table is updated as follows for each date:

update RoomTypeAvailability set numAvailable = numAvailable-1 where roomType = t && date = d

As numAvailable is a primary key, concurrent updates to the table will not be allowed.

Atomically, the Reservation table is also updated in the same transaction.

• To avoid double payment (user clicks on "Pay" twice), a unique identifier (nonce) is associated with each order. A subsequent request to pay will use the same nonce and hence will be identified as duplicate request. This will avoid double-charging the user.

• Recommendation system can use Collaborative or Clustering Algorithms or a combination.

• Each component is stateless and works independently so new instances can be deployed as required, especially during peak time.
• Sharding using Hotel Id can be performed to avoid making Database the bottelneck.
• Content Distribution Network (CDN) is deployed at the edge to serve static content when users are browsing and/or viewing rooms.

Trade offs/Tech choices

Explain any trade offs you have made and why you made certain tech choices...

• Optimistic Locking is used for reservation instead of Pessimistic Locking. Optimistic Locking is more performant as it does not lock any DB rows while reservation is being made. So other users can simultaneously browse and even book a room. Only if there is a conflict, the later transaction is aborted. This also avoids deadlocks.
• Every service is an independent microservice. However, Rooms and Reservation are unified and not independent Reservation needs to update RoomTypeAvailability table also. Keeping them together avoids inter-process communication between them.

Failure scenarios/bottlenecks

Try to discuss as many failure scenarios/bottlenecks as possible.

• If multiple concurrent users are trying to book the same room, only 1 user will succeed at a time due to Optimistic Locking.

• To avoid double payment (user clicks on "Pay" twice), a unique identifier (nonce) is associated with each order. A subsequent request to pay will use the same nonce and hence will be identified as duplicate request. This will avoid double-charging the user.

• A service discovery component like Zookeeper can monitor the health of each microservice and spawn new or more instances as required based on metrics.
• Sharding using Hotel Id can be performed to avoid making Database the bottelneck.
• Load-balancer like NGINX to avoid overloading the API Server

Future improvements

What are some future improvements you would make? How would you mitigate the failure scenario(s) you described above?

• Allow comparison of hotel rates
• Discounts for new and long-time users