Chapter 16 checkpoint

Chapter 16 checkpoint: scaling on AWS

Every question from this chapter’s lessons. Your first answer to each is the one that counts; afterward you can practice again as often as you like. Passing needs 75% right on the first attempt.

0 of 18 answered0 right on the first attempt

Question 1 of 18

The primer’s service gets 100 billion reads a month. Using its rule of 2.5 million seconds in a month, how many reads per second is that on average?

Your answerreads/s

Answers within 5% count.

Question 2 of 18

Put the primer’s scaling steps in order, from one box to millions of users.

Move static content to S3 and MySQL to RDS
Add a memory cache and read replicas
Add a load balancer, several web servers across zones, and a CDN
Add Auto Scaling groups
Move history to a warehouse, some data to NoSQL, and slow work to queues
Run the web server and MySQL on one box

Question 3 of 18

Reads are overloading MySQL. What does the primer say to try before adding a separate memory cache such as ElastiCache?

Tune MySQL’s own cacheMove everything to DynamoDBAdd more web serversShard MySQL

Question 4 of 18

Why does the primer move session data from the web servers into the memory cache?

To analyze sessions in RedshiftBecause sessions are secretSo the web servers are stateless and can be scaled automaticallySo the CDN can serve sessions

Question 5 of 18

An Application Load Balancer uses the default health checks: every 30 seconds, with 2 consecutive failures marking a target unhealthy. About how long after a zone fails does it stop sending that zone’s targets requests?

Your answer

Answers within 10% count, in any unit you choose.

Question 6 of 18

The load needs 30 instances, spread over 3 zones. How many instances must each zone run so that losing any one zone still leaves enough, with nothing new launched?

Your answerinstances per zone

Answers within 1% count.

Question 7 of 18

Aurora keeps 6 copies of each segment, 2 per zone in 3 zones, with a write quorum of 4 and a read quorum of 3. A whole zone fails, and then one more storage node. What can the volume still do?

It still accepts reads and writesIt accepts writes but not readsIt can do nothing until the zone returnsIt can be read, and rebuild copies, but cannot accept writes

Question 8 of 18

Which Auto Scaling policy does AWS recommend when capacity should follow a metric in proportion?

Simple scaling with a 300-second cooldownManual scalingScheduled scaling onlyTarget tracking

Question 9 of 18

Your service runs in three AWS Regions. Which Route 53 routing policy sends each user to the Region with the lowest latency for them?

Geoproximity routingWeighted routingLatency routingSimple routing

Question 10 of 18

Route 53 health checks run every 30 seconds with a failure threshold of 3, and the record’s TTL is 60 seconds. In the worst case, how long can a client keep reaching a failed Region?

Your answer

Answers within 2% count, in any unit you choose.

Question 11 of 18

An Aurora global database takes 400 writes a second, and replication to the secondary Region lags by about 1 second. About how many writes are at risk in an unplanned failover?

Your answerwrites

Answers within 5% count.

Question 12 of 18

Match each disaster recovery strategy to what runs in the recovery Region.

Backup and restorePilot lightWarm standbyMulti-site active/active

Question 13 of 18

Is an Aurora global database active/active for writes?

Each record is written in its home RegionYes, with synchronous replication between RegionsNo: one primary Region takes all writes, and secondaries are read-onlyYes: every Region accepts writes independently

Question 14 of 18

You keep 100 TB of logs in S3 Standard at 0.023 USD per GB-month. A lifecycle rule moves 80% to Glacier Instant Retrieval at 0.004 USD per GB-month and leaves 20% in Standard. What is the new monthly storage bill, ignoring request and retrieval charges, in USD?

Your answerUSD

Answers within 5% count.

Question 15 of 18

Instances in a private subnet send a lot of data to S3 through a NAT gateway. How does a gateway VPC endpoint for S3 reduce the cost?

It compresses objects before uploadIt caches S3 objects on the instancesIt moves objects to a cheaper storage classS3 traffic goes through the endpoint instead of the NAT gateway, and the endpoint has no hourly or processing charge

Question 16 of 18

At the AWS Ohio example rate of 0.045 USD per GB processed, what do 10,000 GB a month of S3 traffic cost in NAT gateway processing charges, in USD?

Your answerUSD

Answers within 2% count.

Question 17 of 18

A team bought a Compute Savings Plan sized for its daytime peak. Its fleet shrinks to a third of that overnight, and the bill is higher than expected. Why?

Usage above the commitment is charged at On-Demand ratesSavings Plans do not apply to EC2They are running in the wrong RegionThe commitment is charged every hour, used or not, so overnight they pay for capacity they do not run

Question 18 of 18

A service runs 10 instances in each of three Availability Zones, each instance handling 1,000 requests a second, and takes 24,000 requests a second. Zone b fails, and replacement instances take several minutes to start. Until they do, which component is the bottleneck?

Application Load BalancerIn all three zones; stops routing to zone b once its health checks fail

Instances in zones a and c20 instances, 1,000 requests/s each

Aurora writer (zone a)Up to 8,000 writes/s; gets 3,000Aurora replica (zone c)Up to 10,000 reads/s; gets 6,000