System's Measurements

Question 1

Latency - Definition

Answer 1

• It’s how long it takes for data to traverse from one point to another point in a system
• There are tradeoffs between latencies of various components

Question 2

Latency - Quantities

Answer 2

• Reading 1 MB sequentially from memory takes approximately 250 microseconds
• Reading 1 MB from SSD takes approximately 1,000 microseconds
• Sending 1 MB of data over a 1 Gbps network takes approximately 10,000 microseconds
• Reading 1 MB from HDD takes approximately 20,000 microseconds
• Sending a packet on a round trip from California to the Netherlands (CA -> NL -> CA) takes approximately 150,000 microseconds

Question 3

Throughput

Answer 3

• How much work a machine can perform in a given time. Typically measured in Gbps, Kbps

Question 4

Availability - Definition and Nines

Answer 4

• Odds of a particular server/service being up and running at any point in time
• Nines:
  - 99% (two 9s): 87.8 hours
  - 99.9% (three 9s): 8.8 hours
  - 99.99%: 52.6 minutes
  - 99.999%: 5.3 minutes. Typically a HA system has this percentage

Question 5

SLA and SLO

Answer 5

SLA (Service Level Agreement):
- Collection of guarantees given to a customer by a service provider
- Typically make guarantees on system’s availability among other things
- SLAs are made up of multiple SLOs

SLO (Service Level Objective):
- A guarantee that’s given to a customer by a service provider
- SLOs constitute an SLA

Question 6

Active and Passive Redundancy

Answer 6

Active redundancy:
- It’s the application of redundant elements at all times. It guards against both system and element failure. For example, the use of multiple LBs to distribute the load in a system
- Allows for element failure, repair, and substitution with minimal disruption of system performance

Passive redundancy:
- It’s the application of redundant elements only when an active element fails. For example, the use of a non-active LB when the active LB fails
- It’s ideal for noncritical elements