Design Mint
Last updated
Last updated
Mint.com is a free, web-based personal financial management service, that allows its users to integrate with financial accounts to automatically extract their data, and manage their personal budget.
Design a system like Mint.
The following are just my own answer. Please read this solution.
Connect Mint account with my financial accounts
Pull my financial data
when I log in
periodically
Give me financial suggestions
userId: GUID of user
bank?: If not provided, sync all my bank info; otherwise, just sync the info of this bank.
Given userId, give a list of suggestions. Example
As we get more and more users, the throughput of the system might not be the first issue, the DB size could be.
We can horizontally scale the DBs.
We can use Consistent Hashing to map the userId to a specific DB instance. When we add or remove DB instance, we just migrate a fraction of the user data to other DB instances.
If the user base continues growing, a single application server can't handle all the requests and we need to scale it.
We can add a load balancer to distribute the load across multiple application servers.
The last thing we hit scaling issue might be the web server since they are only hosting static contents.
We can introduce CDN to improve the availability and performance of the static content retrieval.
We can also scale the single web server to a cluster and add load balaner to distribute the requests.
User: userId, authToken
UserBank: userId, bank, other user data specific to the bank
I'd use SQL DB for User.
Since the UserBank data might be different across different bank. One solution is that we use different tables for different banks. But even if we do that, a specific bank might change their schema as well. So since the data schema might not be stable, I'd prefer to use NoSQL DB.
The relationship between User and UserBank table is connected using userId.
Assume we have 1M DAU, each of them logs in to the portal once per day, then the average QPS is 1M / 24 / 60 / 60 = 11.6 Q/s. If each request taks 100KB, then it's 1.2MB/s.
Assume we have 50M users in total and each user has 3 banks on average, each of them require 2MB information. Then the total storage is 50M * 3 * 2MB = 30TB.
