Misc.

Question 1

What are some key metrics that you should know about caching (specifically around latency, operations per second, and memory bounds)?

Answer 1

For most modern caching systems:
- latency (roughly ~1ms latency)
- operations per second (roughly 100k/second)
- memory bounds (up to ~1TB)

Question 2

What are some signs/triggers that you should may need to consider scaling a cache?

Answer 2

Generally speaking:
- poor hit rate (< 80%)
- higher latency (> 1ms)
- high memory (> 80% usage)
- churn/thrashing

Question 3

What are some key metrics that you should know about most modern databases (specifically around transactions, latency, and storage capacities)?

Answer 3

Most modern databases support:
- high transactions (> 50k/second)
- low-latency (<5ms/read)
- high storage capacities (64TB+)

Question 4

When should you consider possibly scaling your database? What are some common signs?

Answer 4

You should consider scaling if you start to see:
- high write throughput (> 10k writes/second)
- higher latencies (> 5ms read latencies)
- geographic distribution (product)

Question 5

What are some key metrics about your average modern application server (specifically in terms of concurrent requests, CPU, and RAM?)

Answer 5

The average web server should be able to support:
- large number of concurrent requests (100k+/second)
- large CPUs (8-64 cores, 2-4Ghz)
- high amounts of RAM (64-128GB minimum, up to 2TB)

Question 6

What are some key identifiers or metrics that may indicate you need to consider scaling an application server?

Answer 6

You should consider scaling during:
- high CPU (> 70% utilization)
- high memory (> 80%)
- latency threatening SLAs
- connections near capacity (>= 100k)

Question 7

What are some key metrics to know about popular message queues (specifically regarding throughput, latency, and storage)?

Answer 7

Your average message queue (e.g, RabbitMQ, Kafka, etc.) should be able to support:
- extremely high throughput (1M+/second/broker)
- low-latency (<5ms end-to-end latency)
- reasonably high storage (~50TB)

Question 8

What type of identifiers may indicate you need to scale your message queues/brokers?

Answer 8

You should consider scaling or adding appropriate resources if:
- extremely high throughput (~800k+/second)
- high partition counts (~200 partitions/cluster)
- growing consumer lag