Good sense of key facts is helpful for our planning and system design.

Here we try to combine a list of key facts in cloud systems.

Please leave me comments to grow the list together.

Similar Posts:

  • Network Latency by ping test

Within one region less than 0.5 ms; Across regions 70+ ms.

Tested in DigitalOcean, on Feb, 2018

Warning: The result can be quite different in your case.

Rebalancing a big couchbase cluster for adding a new node.

It takes 2 hours, if all nodes in one region.
It takes 10 hours, if all nodes in one region except the new one.

  • Running ‘rm -rf *’ for a 7 TB drive in public cloud, it takes 4+ hours.
Tested in DigitalOcean, on Dec, 2017

Eventual consistency takes time for reclaiming disk capacity.

It might depends on how many files we have, filesystem type, etc.

  • Complete DNS Resolution may take upto 48 Hours.

Question: How you can lower the impact for an uncomfortable/new major DNS change?

  • RDS/Mysql read/write QPS

From this link, we know 1 RDS with db.r3.8xlarge can provide around 7000 QPS.

db.r3.8xlarge: 32 vcpu, 244 GB RAM.

  • RDS/Mysql: avoid storing more than 10 million records in one table

  • Humans can detect ~10ms lag, will tolerate ~100ms
SF--Denver: 50ms
SF--Tokyo: 100 ms
SF--Madrid: 200 ms

  • Availability
Amazon S3: 99.9%
Amazon EC2: 99.95%

Original URL:

Connect with Denny In LinkedIn Or Slack

Share It, If You Like It.

Leave a Reply

Your email address will not be published. Required fields are marked *