Skip to main content

EC2 Associate Level

note

In Stephane Marek's course, this chapter was calling out "associate level" EC2 details.

info

These were the topics I created flashcards for (Remnote) and would revise them using spaced repetition. The formatting is an export from Remnote.

  • Public vs Private IP
    • Public
      • unique in public
      • easily moved specific geo
    • Private
      • unique in private network
      • can be same in different private networks
      • instances need NAT + IG to reach public internet
  • Elastic IP
    • is a public IP
    • provides a fixed IP for your instance
    • restart may change public IP
    • attached to single instance
    • 5 Elastic IP per account (can request more)
    • Best Practice
      • Avoid using EIP―Use random IP + register DNS, or LB instead
  • Placement Groups
    • What are the 3 Placement Group "strategies"? ↓
      • Cluster
      • Partition
      • Spread
    • What's the goal of Placement Groups?―Spread instances over different underlying hardware to minimise risk of concurrent failures.
    • Cluster
      • all instances
        • single AZ
        • single rack
      • low-latency network performance (10gb)
      • riskiest due to no {{redundancy}} if single rack or AZ {{fails}}
      • what kind of applications is a cluster placement group good for?―High Performance Computing (HPC)
    • Partition
      • instances
        • pg-partition
        • multi-AZ in single Region
          • can do 2 x partitions in 1 x AZ and 1 x in another
        • 7 partitions per AZ
        • partitions don't share racks with other partitions
      • what popular DBMS are ideal to use 'Partition' placement groups (hint: H H C K)?―HDFS, HBase, Cassandra, Kafka
    • Spread
      • instances
        • pg-spread
        • separate hardware
        • multi-AZ
          • 7 x VM per AZ per Placement Group
      • use case: {{critical}} applications reduce risk of {{simultaneous}} failures.
  • Elastic Network Interface (ENI)
    • possible attributes
      • primary (private) IPv4 + secondary IPv4
      • 1 x EIP per private IPv4
      • can move them from Ec2 to Ec2
      • bound to AZ
      • elastic network
  • EC2 Hibernate
    • RAM state preserved
    • state written to root EBS volume
    • root EBS volume must be encrypted
    • AMI must be encrypted
    • hibernate is good for...?― ↓
      • long-running process
      • speed up instance initialise time because its not "starting up" from scratch, its loading from point-in-time state.
    • RAM size must be < 150GiB
    • Available for which Instance Types?―On-Demand, Reserved and Spot instances
  • EC2 Nitro
    • better performance "similar to bare metal"
    • High-speed EBS
      • Nitro 64,000 EBS IOPS
      • Non-Nitro max 32,000
    • better (underlying) security
  • EC2 vCPU
    • multi-threading
    • each thread = vCPU
    • maths: 4 x CPU running 2 threads per CPU = 8 vCPUs
    • Options
      • decrease # of CPU cores to decrease license costs
      • turn off multi-threading good for HPC workloads
  • EC2 Capacity Reservations
    • long-term commitment not required (1 & 3)
    • capacity access & billing are immediate
    • required specs ↓
      • AZ
      • number of instances
      • instance attributes e.g. type, tenancy, OS
    • cost savings if combined with RI's and Savings Plans