2.2 Back of the Envelope

Question 2

What is the approximate value of 2^10?

Answer 2

1,024

1,024 is also referred to as 1 KB (kilobyte).

Question 3

What is the exact value of 2^20?

Answer 3

1,048,576

1,048,576 is also known as 1 MB (megabyte).

Question 4

What is the approximate memory size of 2^30?

Answer 4

1 GB

1 GB (gigabyte) is equivalent to 1 billion bytes.

Question 5

What is the latency of an L1 cache reference?

Answer 5

0.5 ns

Question 6

True or False: The latency of a mutex lock/unlock is 25 ns.

Answer 6

False

The latency of a mutex lock/unlock is actually 100 ns.

Question 7

What is the latency for reading 1 MB sequentially from SSD?

Answer 7

1,000,000 ns (1 ms)

This reflects a speed of approximately 1GB/sec for SSD.

Question 8

What happens during a typical first year for a new cluster?

Answer 8

Various hardware failures and maintenance events

Includes overheating, PDU failures, rack-moves, network rewiring, and individual machine failures.

Question 9

What is a key characteristic of Google’s computing environment?

Answer 9

Large clusters of commodity PCs

Question 10

What are the challenges faced in the Google engineering environment?

Answer 10

Need for distributed systems, high capacity systems, and careful design

This is due to data or request volumes that exceed the capability of a single machine.

Question 11

What is the significance of designing software interfaces carefully?

Answer 11

Ensures flexibility and usability for other hypothetical clients

Question 12

What is Protocol Buffers?

Answer 12

A good protocol description language used by Google

It is self-describing, supports multiple languages, and allows for efficient encoding/decoding.

Question 13

What does ‘self-describing’ refer to in the context of protocols?

Answer 13

The ability of a protocol to provide its own structure and meaning

Question 14

What is the estimated time to generate an image results page with 30 thumbnails using a serial read design?

Answer 14

560 ms

Question 15

What is the estimated time to generate an image results page with 30 thumbnails using a parallel read design?

Answer 15

18 ms

This is a theoretical estimate that ignores variance.

Question 16

What is the expected time to quicksort 1 GB of 4 byte numbers?

Answer 16

Approximately 30 seconds

Question 17

What are microbenchmarks used for?

Answer 17

To understand performance and reduce cycle time for performance testing

Question 18

What is the latency for reading 1 MB sequentially from disk?

Answer 18

30,000,000 ns (30 ms)

Question 19

What does the term ‘back-of-the-envelope’ calculations refer to?

Answer 19

Quick estimates to assess feasibility or performance

Question 20

Fill in the blank: The latency for sending a packet from CA to the Netherlands is _______.

Answer 20

150,000,000 ns

Question 21

What are the core language libraries and data structures mentioned?

Answer 21

Core language libraries, basic data structures, SSTables, protocol buffers, GFS, BigTable, indexing systems, MySQL, MapReduce

These are fundamental components for building software systems.

Question 22

Why is it important to understand the implementations of core libraries?

Answer 22

Understanding implementations enables effective back-of-the-envelope calculations

Without this knowledge, one cannot make informed decisions or estimates.

Question 23

What should you do when building infrastructure?

Answer 23

Identify common problems and build software systems to address them in a general way

Avoid trying to satisfy every client demand to maintain system simplicity.

Question 24

True or False: Building infrastructure should be done for its own sake.

Answer 24

False

Infrastructure should be built to address real needs, not hypothetical ones.

Question 25

What does 'Design for Growth' entail?

Answer 25

Anticipate how requirements will evolve and ensure design works if scale changes by 10X or 20X ## Footnote Considerations for larger scales may differ from smaller ones.

Question 26

What is a key goal when designing for low latency?

Answer 26

Aim for low average times and consider 90%ile and 99%ile latencies ## Footnote User experience can significantly degrade with high latency.

Question 27

What consistency model is most products with mutable state gravitating towards?

Answer 27

Eventual consistency ## Footnote This model is better from an availability standpoint compared to strong consistency.

Question 28

What is the significance of using threads in applications?

Answer 28

Using threads can help improve both throughput and latency ## Footnote Modern machines have multiple cores, making parallelization crucial.

Question 29

What should you consider regarding data access?

Answer 29

Disks: seeks, sequential reads; Memory: caches, branch predictors ## Footnote Understanding these aspects is crucial for optimizing performance.

Question 30

Fill in the blank: Compression is a very important aspect of many systems, especially in _______.

Answer 30

inverted index posting list formats, storage systems for persistent data ## Footnote Compression helps reduce storage requirements and improve performance.

Question 31

What are 'canary requests' used for?

Answer 31

To detect issues in applications and improve robustness against failures ## Footnote They help identify problems before affecting a large number of users.

Question 32

What is essential for monitoring and debugging applications?

Answer 32

Exporting HTML-based status pages and key-value pairs via a standard interface ## Footnote This allows for easy diagnosis and monitoring of system health.

Question 33

What is the source code philosophy at Google?

Answer 33

Google has one large shared source base with core libraries and application-specific code ## Footnote This promotes code reuse and improvements across applications.

Question 34

What are some benefits of code reviews and design reviews?

Answer 34

Improve code quality, catch issues early, and facilitate knowledge sharing ## Footnote They are part of good software engineering hygiene.

Question 35

What challenges arise from multi-site software engineering?

Answer 35

Increased coordination needs, communication difficulties, and trust establishment between teams ## Footnote These challenges necessitate effective collaboration tools and practices.

Question 36

What motivates Google's expansion to multiple engineering sites?

Answer 36

To hire the best candidates regardless of geographic location ## Footnote This strategy enhances talent acquisition and diversity.

Question 37

What is a service-based model for software development?

Answer 37

A model that allows fluid changes, easy testing, and enables small teams to accomplish a lot ## Footnote It promotes agility and responsiveness in software development.

Question 38

What impact does work at Google have?

Answer 38

Work has a very large impact, affecting hundreds of millions of users every month ## Footnote This scale of influence can be highly motivating for engineers.