Storage

Scale databases on-demand with Db2 on Cloud

Share this post:

IBM Db2 on Cloud (formerly dashDB for Transactions) has released a scalable plan that allows customers to scale RAM/CPU and disk independently.

Getting started is easy, since it can be deployed and scaled using IBM Bluemix.

Scale your database with sliderbars.

Scaling is easy with Db2 on Cloud. Just use slider bars.

Here are 4 great features of the new plan:

Now that I can scale a database with slider bars, how can I use it?

Db2 on Cloud Flex lets you scale quickly and without complex cloud lingo. So, here are some examples of things you can do:

  • Scale up CPU and RAM before running heavy reports, then scale back down.
  • Scale down resources over the weekend to save costs.
  • Scale down resources for most development work, and then, when running heavy test cases, scale up to production levels.

How long does it take to scale and what about downtime?

Scaling up or down CPU & RAM: 

Time to complete operation: Typically under 20 minutes.

Downtime of scaling event: If your application has appropriate retry logic and you’re using the HA plan, you will not experience any downtime effects. (Click here for detailed instructions.)

Regardless, in 99{07c2b926d154bd5dc241f595a572d3349d41d98f2484798a4a616f4fafe1ebc0} of the cases, an HA plan requires less than 20 seconds of unavailability to scale RAM/CPU. If you don’t have an HA plan, then your system will be down for the full time period, about 20 minutes. HA plans use rolling failover to prevent this.

Details: Scaling CPU up or down on an HA plan should only require a matter of seconds of downtime. However, if you use large transactions, they may need to be rolled back if interrupted by a scaling event. This is the situation that can cause a longer delay. If you plan to scale often, you can commit transactions more frequently to help prevent this.

Scaling up storage space:

Time to complete operation: If your total database storage is under 100 GB, then scaling up storage is a fully online operation. Beyond that 100GB point, typically under 20 minutes.

Downtime of scaling event: Again, a database that’s under 100 GB will scale fully online. Beyond that, an HA database may require up to 10 minutes of downtime to scale up storage. A non-HA plan may require up to 20 minutes.

Details: How do I know if my storage scaling will require interruption in service? Our cloud service is designed to handle most storage scaling online. But at certain storage levels, extra operations are needed to scale up in a performant manner, which will require the offline time stated above. These are: 100 GB, 1 TB, 2 TB, 3 TB, 4 TB. As long as you’re scaling between those amounts, the storage scaling will be transparent.

As always, IBM is striving to move more operations to function fully online, so stay up-to-date with the latest news on our scaling capabilities.

What is the pricing?

The pricing below is accurate as of August, 2017. Be sure to check the Bluemix Db2 on Cloud tile for any pricing updates.

The pricing is based on consumption:

  • A starting price of $189 USD per month per node.
  • Pay $13 per month for each additional GB of RAM. (Or, equivalently, $52 per CPU core)
  • Pay $1 per month for each additional GB of storage space.
  • Pay an additional $0.20 per 1 million IO operations that you consume

Keep in mind, HA plans will require 2x the cost for nodes, RAM and storage.

Example: An HA system with 2 cores and 8 GB of RAM will cost $482 USD/month:

Nodes: $189 x 2 (regular node & HA node) = $378

RAM/Cores: $13 x 4 (8 GB RAM – 4 GB RAM free with base system) X 2 for HA = $104

Total: $482, (plus IO fees at $0.20 per 1 million IO operations).

Scaling via Command Line

Scaling via command line or automation is as simple as running one line of cloud foundry code:

Basic example:
cf update-service my_flex_service -c '{"ram_gb":8,"email":"myemail@mydomain.com”,”scale_time":0}'

Notice that you can include an email address right inside the command for alerts. You can send alerts to multiple emails separating them with a comma.
This saves you the hassles of managing any email code. (This is handy, because dealing with email code can get icky– you’d need to deal with yet more APIs, SMTP problems or with securely storing credentials for your email API service.)

More Storage stories
April 25, 2019

What is Object Storage?

In our latest lightboarding video, Anirup Dutta explains how object storage works, lists some of the benefits, and give you some use cases for when object storage may be your best option.

Continue reading