AIAI Agent Dev Labaiagentdev-lab.com
Sign in to save progressSign in
Menu

SAP-C02 · LESSON 4

Accelerate Workload Migration and Modernization20% · 4 topics

Learn to select migration candidates, choose migration approaches and new architectures, and identify modernization opportunities.

4.1Select existing workloads and processes for potential migration

4.1 Select existing workloads and processes for potential migration.

Concept

To select migration candidates, first understand the current state. Collect server configuration, utilization and dependencies with AWS Application Discovery Service (agent-based or agentless), and track progress centrally with Migration Hub. Group strongly dependent servers into the same migration group (wave), and order waves by business priority and risk.

How it works / steps

  1. Collect server information and network dependencies with Discovery Service
  2. Group servers by application
  3. Prioritize by business value, complexity and risk
  4. Plan migration waves considering dependencies
  5. Track progress with Migration Hub

Workplace example

Dependency visualization shows the order system and inventory DB communicate heavily, so migrate both in the same wave to avoid latency problems.

Common confusions and exam tips

Mini quiz

Q1. For migration planning, you want to collect configuration and utilization of 500 VMs on on-premises VMware without installing agents on each server. Which method is most appropriate?

  1. Create only a cost estimate with the Migration Evaluator collector
  2. Collect each server's configuration with Systems Manager Inventory
  3. Install the Application Discovery Service agent on every server
  4. Connect the Application Discovery Service agentless collector to vCenter
Show answer and explanation

Answer: D Connect the Application Discovery Service agentless collector to vCenter

The agentless approach of Application Discovery Service collects VM configuration and utilization from vCenter. Use agents if you need detailed process and network dependencies. [Other options] Agents break the no-agent condition. A cost estimate alone lacks the configuration and utilization needed. Systems Manager Inventory needs an agent on each server.

Q2. You are planning the migration of hundreds of apps in waves. Which approach is most appropriate?

  1. Migrate all servers at once on the same weekend, finishing in one go without waves
  2. Divide waves in order of server name and proceed mechanically without regard to dependencies
  3. Move the most important and complex core system in the first wave to tackle the hardest part while experience is still low
  4. Group strongly dependent servers in the same wave, and move low-risk apps first to solidify the procedure
Show answer and explanation

Answer: D Group strongly dependent servers in the same wave, and move low-risk apps first to solidify the procedure

Splitting dependencies increases traffic between on-premises and the cloud, making performance and failure problems more likely. Start small, establish procedures, then scale up.

Reference links

4.2Determine the optimal migration approach for existing workloads

4.2 Determine the optimal migration approach for existing workloads.

Concept

Choose a migration strategy (the 7 Rs: rehost, replatform, refactor, repurchase, retire, retain, relocate) and decide on tools. Rehost servers with AWS Application Migration Service (MGN: continuous block-level replication with short cutover), databases with AWS DMS (full load and CDC for minimal downtime) and Schema Conversion, files with DataSync, and large data over thin networks with Snowball Edge.

How it works / steps

  1. Decide a 7 Rs strategy for each workload
  2. Replicate servers with MGN and cut over after test launches
  3. For DBs, sync changes with DMS CDC after a full load, and cut over with a short outage
  4. When moving to a different DB engine, also convert the schema
  5. Decide online or offline (Snowball) from line bandwidth and deadline

Workplace example

100 TB of files must be moved in one week over a 1 Gbps line (the line cannot make it), so ship the initial copy with Snowball Edge and sync changes with DataSync.

Common confusions and exam tips

Mini quiz

Q1. During migration with DMS, CDC latency keeps increasing and cannot catch up. What is the most appropriate next step?

  1. Switch LOB columns to full LOB mode to prevent missing data
  2. Enlarge the target DB instance to speed up writes
  3. Restart the task from full load to reset the CDC start position
  4. Check the replication instance's CPU, memory and more, and move to a larger class if needed
Show answer and explanation

Answer: D Check the replication instance's CPU, memory and more, and move to a larger class if needed

CDC latency can be caused by insufficient replication instance resources, target write performance, task settings and so on. Identify the cause with metrics and address it. [Other options] Restarting takes time without finding the cause. Target performance can also be a cause, but check metrics first. Full LOB mode tends to be slower.

Q2. Which statement about AWS Snowball Edge is incorrect?

  1. Data on the device is not encrypted, so users must protect data in transit by other means
  2. Data written to the device is imported into S3 after it arrives at AWS
  3. Large amounts of data that would take too long over the network can be migrated by physically shipping a device
  4. It also has features usable for data processing (compute) at the edge
Show answer and explanation

Answer: A Data on the device is not encrypted, so users must protect data in transit by other means

Data written to Snowball Edge is automatically encrypted with keys managed in KMS, and the device has tamper-resistant mechanisms.

Reference links

4.3Determine a new architecture for existing workloads

4.3 Determine a new architecture for existing workloads.

Concept

For the target architecture, choose compute (EC2, containers, serverless), storage (EBS, EFS, FSx, S3), databases (RDS, Aurora, DynamoDB, purpose-built DBs), networking and authentication from requirements. Also consider FSx for Windows File Server for Windows file shares, commercial DB licensing constraints and Active Directory integration (AWS Managed Microsoft AD).

How it works / steps

  1. Organize current functional, performance and licensing requirements
  2. Choose parts that can be replaced with managed services (DB, files, AD)
  3. Decide license-included or BYOL (Dedicated Hosts) from license terms
  4. Design networking and authentication for the period of coexistence with on-premises
  5. Design post-migration operations (monitoring, backup)

Workplace example

Move Windows file servers to FSx for Windows File Server, and integrate Active Directory through a trust between AWS Managed Microsoft AD and on-premises AD.

Common confusions and exam tips

Mini quiz

Q1. The investigation found that apps being migrated depend on on-premises Active Directory authentication. AD will stay on-premises for now. What is the most appropriate next step?

  1. Create a trust between AWS Managed Microsoft AD and on-premises AD, or relay with AD Connector
  2. Sync on-premises AD users to IAM users with IAM Identity Center
  3. Move AD users to an Amazon Cognito user pool
  4. Create Simple AD and import on-premises AD users
Show answer and explanation

Answer: A Create a trust between AWS Managed Microsoft AD and on-premises AD, or relay with AD Connector

With Directory Service, establish a trust between AD on AWS and on-premises AD, or relay authentication to on-premises AD with AD Connector. [Other options] Identity Center manages sign-in to AWS and doesn't replace the app's AD authentication. Simple AD can't trust on-premises AD. Cognito can't run AD-dependent apps as is.

Q2. You are choosing the runtime environment for an app being migrated. The app needs specific kernel modules and a license management agent and requires fine-grained control of the OS. Which is the most suitable?

  1. AWS Elastic Beanstalk
  2. AWS Lambda
  3. Amazon EC2
  4. Amazon ECS on AWS Fargate
Show answer and explanation

Answer: C Amazon EC2

EC2 gives OS administrator access, so apps needing special drivers or agents can run. In exchange, users handle operations such as patching.

Reference links

4.4Determine opportunities for modernization and enhancements

4.4 Determine opportunities for modernization and enhancements.

Concept

Modernization makes migrated systems cloud-native to improve agility and cost. Options include splitting monoliths into microservices (the strangler fig pattern), containerizing (ECS, EKS, Fargate), going serverless (Lambda, Step Functions, EventBridge), moving from commercial DBs to open source or Aurora, and adopting purpose-built DBs.

How it works / steps

  1. Identify parts with the biggest operational burden, change frequency or cost issues
  2. Carve out functions gradually with the strangler fig pattern
  3. Reduce operational effort by containerizing or going serverless
  4. Move commercial DBs to Aurora or similar to lower license costs
  5. Measure the effect (deployment frequency, cost)

Workplace example

From a monolithic e-commerce site, first carve out product search into API Gateway and Lambda, route to it by ALB path, and migrate gradually while leaving the rest of the monolith unchanged.

Common confusions and exam tips

Mini quiz

Q1. You decided to gradually move a monolithic order system to microservices. What is the most appropriate first step?

  1. Install every tool needed for microservices so they are available even with no planned use
  2. Split the monolith's database by every function first and make all app changes together at the end
  3. Rebuild all functions as microservices at once and switch everything over when complete
  4. Pick one low-impact function, carve it out as a new service and route to it by ALB path
Show answer and explanation

Answer: D Pick one low-impact function, carve it out as a new service and route to it by ALB path

Carve out something small, confirm the effect and operating procedures, then expand scope (strangler fig). Rebuilding everything at once is high risk.

Q2. You are considering AWS Lambda as a migration target. Which statement about Lambda is incorrect?

  1. It runs code triggered by events and charges by execution time and number of requests
  2. A single invocation can run for any number of hours, making it ideal for batch jobs that take several hours
  3. Functions can also be deployed as container images
  4. Concurrency scales automatically with load
Show answer and explanation

Answer: B A single invocation can run for any number of hours, making it ideal for batch jobs that take several hours

A single Lambda invocation runs for at most 15 minutes. For long processing, split it with Step Functions or use Fargate, AWS Batch and so on.

Reference links

Lessons are written by this site based on public information and do not guarantee exam content. Product specifications change, so check the latest information at the reference links.

Notes (unofficial, original questions, no pass guarantee)