Method challenge: estimate, then design
The situation
A social app knows its usage, not its rate. Each page is put together for its viewer from posts that many people read. Work out the peak's requests a second from the usage figures below, as step 1 of the method does, before you design for it.
Traffic
2.16 million people a day. Each makes 60 page views (reads) and 3 posts (writes). The peak runs at 5 times the day’s average rate.
| Operation | Steps |
|---|---|
| Read | Read data |
| Write | Write data |
- ReadRead data
- WriteWrite data
Data
A cache answers 85% of the reads that reach it.
What you are proving
Usage figures turned into the right rate: the design carries the evening peak, not only the day's average, and every tier is sized for it.
Rules
- A CDN cannot answer these requests.Every page is put together for the person viewing it, so no two users get the same response.
- Posts stay in the relational database, and every write is stored in it.The posts are already there.
Builds on
Targets
- Serve every request at the peak
- At least 99.9% available, with no single point of failure
- At most 2,050 USD a month
- Every write kept in a relational database
- Any tier loses one instance: nothing dropped, no tier over 80% busy