Task 3.3: Determine a strategy to improve performance.

• APut an AWS Global Accelerator standard accelerator in front of the game servers in both Regions

  Correct

  Global Accelerator provides static anycast IP addresses and improves TCP and UDP applications by carrying their traffic over the AWS global network to endpoints in the nearest Region.

• BServe the game traffic through an Amazon CloudFront distribution with the game servers as origins

  Incorrect

  CloudFront speeds up cacheable and dynamic web content. For non-HTTP use cases such as UDP gaming, AWS points to Global Accelerator instead.

• CUse Route 53 latency-based records that point players to the Region with the lowest latency

  Incorrect

  Latency records pick a Region at DNS time, but traffic still crosses the public internet, and the addresses are not a small fixed list.

• DMove the game servers to larger instances with enhanced networking in both Regions

  Incorrect

  Enhanced networking speeds up the instance network interface. It does not shorten the internet path that causes the jitter.

Non-HTTP traffic, a global audience and fixed IP addresses point to Global Accelerator.

• AAlarm on the Maximum statistic of TargetResponseTime above 300 ms

  Incorrect

  Maximum reacts to a single slow request, so it would fire even when 99.9% of requests are fast and the contract is met.

• BAlarm on the p99.9 percentile of TargetResponseTime above 300 ms

  Correct

  Percentile statistics show the value below which a given share of data points fall, so p99.9 measures the contract directly; averages hide slow requests.

• CAlarm on the SampleCount statistic of TargetResponseTime above a fixed number

  Incorrect

  SampleCount counts requests. It says nothing about how long they took.

• DLower the Average threshold from 300 ms to 100 ms so that the alarm fires earlier

  Incorrect

  A lower average threshold still hides the slowest requests and is not the measure the contract uses.

Translate the SLA into the same statistic: a percentile target needs a percentile alarm.

• AMove the web servers to the next larger instance size and raise the Auto Scaling group's maximum size

  Incorrect

  Bigger servers add capacity but still serve every file from one Region, so distant readers stay slow and the repeated work remains.

• BAdd an Amazon ElastiCache cluster that the web servers use to cache the static files in memory

  Incorrect

  A cache next to the servers saves some disk reads, but every byte still travels from the Region to distant readers.

• CPut AWS Global Accelerator in front of the load balancer to route readers over the AWS network

  Incorrect

  Global Accelerator improves the network path but does not cache content, so the servers would still serve every file.

• DServe the site through Amazon CloudFront, with the static files moved to Amazon S3 as a second origin

  Correct

  CloudFront caches content at edge locations close to readers, which cuts latency and takes repeated requests off the origin.

Repeated static content for a global audience belongs at the edge: S3 as the store and CloudFront as the cache.

• ATurn on EBS encryption for the volume with a customer managed KMS key

  Incorrect

  Encryption protects the data at rest. It does not add IOPS or bandwidth.

• BChange the volume to Throughput Optimized HDD (st1) for higher throughput

  Incorrect

  HDD volumes suit large sequential I/O. For random database I/O they deliver far fewer IOPS than SSD volumes.

• CModify the volume to gp3 and provision the IOPS and throughput that the peak needs

  Correct

  gp3 lets you provision IOPS and throughput independently of volume size, so the volume no longer depends on the size-based baseline of gp2.

• DMove to an instance type with more dedicated EBS bandwidth and IOPS

  Correct

  Each EBS-optimized instance type has a maximum EBS bandwidth and IOPS; if the instance is at that limit, faster volumes alone cannot help.

• ETake EBS snapshots more often so that each incremental snapshot is smaller

  Incorrect

  Snapshot frequency affects backups, not the volume's performance during peak load.

A queue on the volume plus an instance at its EBS limit means both sides need more: provisioned gp3 performance and an instance with more EBS bandwidth.

• AField-level encryption for viewer requests

  Incorrect

  Field-level encryption protects data; it does not reduce origin load.

• BCloudFront Origin Shield

  Correct

  Origin Shield is an additional caching layer that helps minimize the origin's load and improve its availability.

• CA shorter cache TTL

  Incorrect

  Shorter TTLs send more requests to the origin.

• DGeographic restrictions by country

  Incorrect

  Blocking countries reduces audience, not cache misses.

An extra central cache layer in front of the origin is Origin Shield.

• ARemove the TTL from all cache entries so they never expire

  Incorrect

  Without a TTL, stale entries stay even longer.

• BRead prices only from read replicas of the database

  Incorrect

  Bypassing the cache removes its benefit and adds database load.

• CWrite price changes to the cache too, and add a TTL

  Correct

  With lazy loading alone, cached data can become stale; write-through and TTL strategies address this.

• DIncrease the cache node size

  Incorrect

  A larger node holds more data but still serves stale entries.

Stale cache data is fixed with write-through updates and a TTL.

• ALaunch the existing instance types into a cluster placement group so that the nodes sit close together in the Availability Zone

  Incorrect

  Cluster placement groups suit low-latency traffic between instances, but AWS notes that the lowest latency and the highest packets per second also require an instance type that supports enhanced networking, which these instances lack.

• BPut an AWS Global Accelerator accelerator in front of the nodes so that their traffic travels over the AWS global network

  Incorrect

  Global Accelerator improves the path from internet clients to applications through AWS edge locations and adds hourly and data transfer charges. It does nothing for traffic between nodes in one zone.

• CAttach a second elastic network interface to every node and split the grid's messages across the two interfaces

  Incorrect

  Multiple network interfaces serve needs such as management networks, network appliances and dual-homed instances. They do not give these instances enhanced networking or lift their packets-per-second ceiling.

• DMove to current-generation instance types of similar size that support ENA, and make sure enhanced networking is turned on

  Correct

  Enhanced networking provides higher bandwidth, higher packets per second and consistently lower latency between instances, at no additional charge. All Nitro-based instances use ENA for it, and the change needs no application code changes.

The constraints are more packets per second, lower latency, no application change and no extra networking charge. A cluster placement group helps only together with enhanced networking, Global Accelerator serves internet clients at a cost, and extra interfaces do not change the instance's network performance. Current-generation instances with ENA meet every constraint.

• AA longer function timeout

  Incorrect

  Timeouts do not prevent cold starts.

• BMoving the function into a VPC

  Incorrect

  VPC access does not keep environments initialized.

• CReserved concurrency set to the peak value during the evening

  Incorrect

  Reserved concurrency guarantees and caps concurrency but does not initialize environments in advance.

• DProvisioned concurrency scheduled for the evening peak

  Correct

  Provisioned concurrency keeps pre-initialized execution environments ready to respond immediately to requests.

Predictable peaks are covered with provisioned concurrency for that time.

• AAWS Trusted Advisor performance checks

  Incorrect

  Trusted Advisor checks accounts, not user page loads.

• BCloudWatch RUM for the web application

  Correct

  CloudWatch RUM collects client-side performance data from actual user sessions in near real time.

• CVPC Flow Logs for the web subnets

  Incorrect

  Flow logs record IP traffic, not page load times.

• DCloudWatch metrics for the load balancer

  Incorrect

  Load balancer metrics are measured on the server side, not in users' browsers.

Real user performance is measured with CloudWatch RUM.