First look: Google Cloud SQL takes on Amazon Aurora

  • Whatsapp
google cloud sql create instance


Google’s first attempt at MySQL as a service, Cloud SQL, launched in 2011, was somewhat of a disappointment from the standpoint of performance and scalability. Cloud SQL Second Generation, which was released on Aug. 16 of this year, has no such problems. 

Bacaan Lainnya

In essence, Cloud SQL Second Generation is an implementation of MySQL 5.7 on top of Google’s Compute Engine and Persistent Disk. According to Google, Cloud SQL Second Generation runs seven times faster and has 20 times more storage capacity than its predecessor — with lower costs, higher scalability, automated backups that can restore your database from any point in time, and 99.95 percent availability, anywhere in the world.

Cloud SQL features

Cloud SQL Second Generation is a fully managed database service, taking the knob-twiddling out of MySQL configuration. As you can see in the figure below, when you configure Cloud SQL you set the storage type, the initial storage capacity, and whether you want a fail-over replica, automated backups, binary logging, and automatic storage capacity expansion.

Creating a Google Cloud SQL instance involves making only the necessary decisions: the storage type, the initial storage capacity, and whether you want a fail-over replica, automated backups, and binary logging. Further down on this form, there is a check box to enable automatic storage capacity expansion.

You don’t see multiple pages of settings: The database has already been optimized for performance. By contrast, Amazon Aurora does show you all of the MySQL 5.6 settings and allows you to change them, but also has well-tuned defaults.

I asked the Google Cloud SQL team how they optimized the database for performance. They answered:

Cloud SQL uses industry-standard best practice tuning parameters to achieve current performance levels. We will continue to optimize MySQL and the operating system to take advantage of our underlying Google Cloud Platform infrastructure. We will also continue to improve and innovate on our infrastructure to improve the product’s performance. Our current configuration is typical of what can be achieved using Compute Engine instances. Some specifics are:

  1. Ext4 filesystem options: noatime, discard
  2. Disk scheduler: set to “noop”
  3. MySQL options:
       
  • innodb_buffer_pool_size is set by machine type (amount of available RAM)
  • innodb_flush_method = O_DIRECT
  • innodb_flush_neighbor_pages = 0
  • innodb_log_file_size = 512MB

The management of the database has been automated as much as makes sense. Backups, replication, patches, and updates are all done for you on your schedule, in a way that ensures better than 99.95 percent availability. That can’t be Five 9’s availability because patches and updates have to happen, but at least they occur only every few months during your chosen maintenance window.

Copyright © 2016 IDG Communications, Inc.



Source

Pos terkait

Tinggalkan Balasan

Alamat email Anda tidak akan dipublikasikan. Ruas yang wajib ditandai *