Updated: Dec 28, 2020
1. (Thousands of thumbnails and write temporary data to storage in the process) Disks that require high-speed IO (high-rate random disk read and write long-term consistent durability): Amazon EBS provisioned IOPS SSD volume
2. Storage: S3, S3 is not suitable for frequent coverage and deletion. The most useful use of RDS is to add, delete, modify and check, so if the data needs to be overwritten and deleted all the time, RDS is required
3. I/O is low, occasionally there will be small peaks: EBS General Purpose SSD (gp2) General Purpose
4. Amazon Athena database query service is a serverless interactive SQL query service that allows you to easily use standard SQL for S3 database query work.
5.“Mount storage” keyword: EFS
EFS is selected for data consistency and file locking
EFS is a file store, not an object store (S3)
6. Throughput optimized HDD (st1) volumes provide low-cost magnetic storage, which defines performance in terms of throughput rather than IOPS. This volume type is very suitable for large continuous workloads such as Amazon EMR, ETL, data warehouse, and log processing.
7. Glacier: The data is immutable and cannot be updated after the archive is created.
8. SNS+SQS combination. SQS queues messages and SNS pushes them wherever they need to go. Can prevent bottlenecks and single points of failure.
9. Mobile/web application authentication, think about Cognito and MFA.
10.gateway VPC endpoint is only used for DynamoDB and S3, the remaining other services are all using interface VPC endpoint.
11.VPC flow logs: record network traffic.
12. Encrypt S3 and use SSE-S3 without managing any keys.
13.Cloudtrail is used to record API calls and can be used to record behavior audit logs of any resource of AWS.
To confirm the reason for the instance termination, you can use the Amazon CloudWatch agent to push logs to Amazon CloudWatch Logs.
14.CloudWatch Logs is used to solve log problems.
15. Elastic Beanstalk will automatically handle the detailed deployment tasks such as capacity pre-configuration, load balancing, automatic expansion, and application health monitoring. (There will still be EC2 instances, but it is automatically managed) is a very good web solution, such as node.js.
16. CloudWatch Metrics will only show you metrics specific to ELB itself. It is not which src IP/client accessed a specific service on the ELB. To do this, you need to enable access logs on ELB.
17. There are multiple app calls and consumption, and they are exposed as RESTful calls, then Kinesis Data Stream is required. SQS is generally used for internal applications.
18. AWS Systems Manager Parameter Store is used for configuration and password management. Those configurations and passwords are used as parameter values. Some services require passwords, so you can call AWS Systems Manager Parameter Store to get the corresponding This is a solution for AWS.
19. If downtime is not acceptable, DynamoDB is more suitable for storing user session data, and it is wirelessly expandable (unlimited scalability) to support terabytes of data.
20. Strictly speaking, session data is the most suitable to be placed in the cache, so choose Elastic Cache.
21. The path-based routing in the request URL is ALB
22. If the instance fails to pass the health check, ALB will be triggered to be deleted from the new instance, which can be used to automatically restore services.
If you attach one or more load balancing or target groups to your elastic scaling group, by default, the group will not consider an instance to be unhealthy, and if the instance fails the load balancing health check, it will be replaced.
However, you can choose to configure the Auto Scaling group to use elastic load balancing health checks. This ensures that the group can determine the health of the instance based on other tests provided by the load balancer. The load balancer periodically sends pings, attempts to connect, or sends requests to test EC2 instances. These tests are called health checks.
23.Reader endpoint only load balances connections to the available Aurora replicas in the Aurora DB cluster.
24. Changing the health check type of the auto-scaling group to ELB can ensure that the instances marked as abnormal will be terminated and replaced. The health check for ELB can detect the number of specific instances, and then Auto scaling will perform the necessary adjustment.
25.write-heavy optimized disk, for example, for RDS/NoSQL, Provisioned IOPS SSD is preferred
read-heavy consider using Elastic Cache.
26. For traffic routing through content metadata (content metadata), consider Application Load Balancer.
27. If the application accesses the S3 object and still reads outdated data, it is possible that the application updates the record by overwriting the existing object with the same key.
28. Generally, EC2 and RDS use the internal network, and the external network passes through ELB and NAT Gateway.
29. An egress-only internet gateway can be used for instances in the private subnet to establish IPV6 communication to the internet.
30.4 iGB = 200 IOPS
Volume(iGB): IOPS = 1:50