[System Design 101] Lesson1 Making Estimates

Q: How to measure your network performance?

A: There are 3 key metrics to measure your network performance: latency, throughput, and bandwidth.

• Latency(time): delay or lag of your data(packet) transmission, e.g. Server A in New York sends a data packet to Server B in London. Server A sends the packet at 05:35:00.000 GMT and Server B receives it at 05:35:00.139 GMT. The amount of latency on this path is the difference between these two times: 0.139 seconds or 139 milliseconds.
• Throughput(volume): actual rate of successful data transfer
• Bandwidth(capacity): maximum theoretical capacity of a network connection, measured in megabytes per second (MBps)

credit: https://www.dnsstuff.com/latency-throughput-bandwidth

Q: What are the system components benchmark for throughput, latency, and bandwidth?

A: Generally speaking:

Latency(speed)

• CPU cache > RAM(Random-access memory) > disk > local network > global network
• The rank is highly related to the distance(the farther the distance the longer the transmission time it takes)

Throughput(volume)

• RDBMS writes/sec are only a half of its reads/sec
• RDBMS and Distributed cache and message queue share same reads/sec
• Distributed cache and message queue have same reads/sec writes/sec
• NoSQL usually has 2x-5x reads and writes than RDBMS

Bandwidth

• NoSQL can be unlimited like using sharding or hashing.

Reference

• https://www.cloudflare.com/learning/performance/glossary/what-is-latency/#:~:text=Suppose%20Server%20A%20in%20New,0.145%20seconds%20or%20145%20milliseconds.
• https://www.udemy.com/course/pragmatic-system-design/learn/lecture/28373740#overview