Skip to main content

High Lights- 3

Launch Templates and Tenancy​

  • A launch template specifies instance configuration (AMI ID, instance type, key pair, security groups, etc.).
  • Default tenancy: shared.
  • Tenancy behavior:
    • Launch Template: shared, VPC: dedicated → dedicated tenancy
    • Launch Template: dedicated, VPC: default → dedicated tenancy

EC2 Tenancy Options​

  • By default, Amazon EC2 instances run on shared tenancy.
  • Dedicated Instances:
    • Run on hardware dedicated to one customer.
    • Isolated across AWS accounts but can share hardware across instances within same account.
  • Dedicated Hosts:
    • Full control over instance placement.
    • Visibility into physical server.

image

EC2 Spot Instances​

  • Spot Instance = unused EC2 at reduced price.
  • Pricing is determined by long-term demand/supply in each AZ.
  • Spot Instance request types:
    • One-time
    • Persistent
  • Persistent requests reopen after interruption.
    • Spot Fleets maintain target capacity by launching replacements.
    • Canceling requests:
      • May or may not terminate instance depending on config.

EC2 Recovery​

  • Recover impaired EC2 instance using CloudWatch alarms.
  • Recovery applies to hardware failure or AWS-repairable issues.
  • Preserves:
    • Instance ID
    • Private & Elastic IPs
    • Metadata
  • If instance is in a placement group, recovery remains within group.
  • Note: Data in memory is lost.

🧠 Amazon EBS Volume Types — Memorize Like a Pro​

✅ Provisioned IOPS SSD (io1) — Best for High IOPS Needs​

  • Backed by: SSD (Solid-State Drives)
  • Use Case: Mission-critical, I/O-intensive workloads (e.g., production-grade databases)
  • IOPS Performance:
    • Up to 50 IOPS/GB
    • Max 64,000 IOPS per volume
  • Throughput: Up to 1,000 MB/s per volume
  • Why Choose: Meets use-case needing 25,000 IOPS

❌ Incorrect Choices (Not suitable for high IOPS needs)​

🟡 General Purpose SSD (gp2) — Balanced for General Use​

  • Backed by: SSD
  • Use Case: Dev/Test, Boot volumes, Web servers
  • Max IOPS: 16,000 IOPS (not sufficient for 25k+)
  • Throughput: Moderate
  • 🧠 Good all-rounder, but not built for intense workloads

🔵 Cold HDD (sc1) — For Archival Storage​

  • Backed by: HDD (Hard Disk Drive)
  • Use Case: Infrequently accessed, cold datasets
  • Max IOPS: 250 IOPS
  • Throughput: Low
  • 🧠 Think: Cheap, cold storage for backups

🟣 Throughput Optimized HDD (st1) — For Big, Streaming Data​

  • Backed by: HDD
  • Use Case: Throughput-intensive workloads (e.g., MapReduce, Kafka, ETL, log processing)
  • Max IOPS: 500 IOPS
  • Throughput: High
  • 🧠 Great for throughput, bad for high IOPS

🔑 Summary Table​

Volume TypeBacked ByMax IOPSBest For
io1SSD64,000High IOPS DBs
gp2SSD16,000Dev/Test
st1HDD500Big data streaming
sc1HDD250Archival

FSx for Lustre​

  • Use FSx for Lustre with Scratch File System for:
    • Temporary storage
    • Short-term data processing
  • No replication, non-persistent if a file server fails.
  • Up to 6× burst throughput of baseline 200 MBps/TiB.

Amazon Redshift​

  • Amazon Redshift is a fully managed petabyte-scale data warehouse in the cloud.
  • Can be a target for data migration from other databases.

Route 53: Alias vs CNAME​

  • Amazon Route 53:
    • No charge for alias queries, but charges for CNAME.
    • Alias → AWS resources only (S3, CloudFront, same hosted zone).
    • CNAME → Any DNS record.
  • At zone apex (covid19survey.com):
    • Cannot create CNAME
    • Can create alias (e.g., point to www.covid19survey.com)

Internet Gateway​

  • Internet Gateway:
    • Acts as route target for internet-bound traffic.
    • Performs network address translation (NAT) for public IPv4 instances.

Amazon SQS Polling​

  • Amazon SQS supports:
    • Short polling: immediate response, may be empty.
    • Long polling: waits for messages up to a timeout.

      Long polling makes it inexpensive to retrieve messages, reducing cost and empty receives.

Amazon SQS FIFO Throughput​

  • By default, FIFO queues support up to 3,000 messages/sec with batching.
  • Without batching: up to 300 messages/sec (send, receive, or delete operations).
  • Batching enables meeting higher throughput requirements (up to 3,000 messages/sec).

AWS CloudFormation StackSets​

  • StackSet = CloudFormation + multi-account + multi-region.
  • Uses one template from admin account.
  • Provisions into target accounts/org units across regions.

Elastic Load Balancer: Connection Draining​

  • Elastic Load Balancing:
    • Uses connection draining for:
      • De-registering or unhealthy instances.
      • Preserves existing connections until timeout (1–3600s, default: 300s).
    • After timeout, force-closes remaining connections.

DynamoDB Global Tables​

  • Active-active configuration in DynamoDB global tables:
    • No failover – data is written regionally and replicated.
    • More expensive than Aurora Global Database.

VPC Sharing​

  • VPC sharing via Resource Access Manager (RAM):
    • Allows multiple accounts to deploy resources in a shared VPC.
    • Owner shares subnets (not VPC itself) with org participants.

Elastic Fabric Adapter (EFA)​

  • Elastic Fabric Adapter (EFA):
    • Enhances HPC and ML inter-instance communication.
    • Adds OS-bypass interface to ENA features.
    • User-space apps can access transport directly for performance.

Elastic Cache​

  • Amazon ElastiCache can be used to significantly improve latency and throughput for many read-heavy application workloads (such as social networking, gaming, media sharing, leaderboard, and Q&A portals) or compute-intensive workloads (such as a recommendation engine) by allowing you to store the objects that are often read in the cache.