Pushing Cloud MySQL Performance The Oracle Way

Managing MySQL databases in the cloud is no easy chore, according to Steve Zivanic, vice president of database and autonomous services at Oracle.

Using Amazon Web Services as an example, Zivanic said that typically, if a company needs a database for transactional data, it will use AWS Aurora. For analytics, it will use Redshift, and to move data between the two, the company will need to leverage AWS Glue or Kinesis Data Firehose. At a time when the amount of data being generated is growing exponentially and organizations are trying to draw as much useful business information from it as quickly as possible, that’s a lot of time moving the data from one database to another.

It also can be costly, from paying for multiple cloud services to the cost of moving the data itself.

In December 2020, Oracle Cloud introduced HeatWave, an in-memory query accelerator for its MySQL Database Service that bolstered MySQL’s capabilities for running analytics workloads. It scales to thousands of cores and, according to Oracle, is 1,100 times faster than Aurora and 2.7 times faster than Redshift and does all this at a third of the cost.

A key to HeatWave is that it enables organizations to leverage Oracle Cloud’s MySQL Database Service for both transactional and analytics tasks, running OLTP [online transactional processing] and OLAP [online analytical processing] workloads from the single service. This reduces the cost, complexity and time that is created when data needs to be moved from a transactional to an analytics database.

“If you can do it in one database and you can do it faster and cheaper, why would not I switch?” Zivanic tells The Next Platform. “This is why we’re having an increasing number of companies that are doing that. We’ve combined basically the ability to OLTP and OLAP. You can run all your real-time analytics against the MySQL database. No ETL [extract, transform and load], no changes to MySQL applications. What we’ve realized – and now many customers are obviously realizing – is real-world applications both have transactions and analytics complex queries. It’s like saying I have one car with an accelerator and another car with brakes. You need them in the same thing.”

That capability is gaining traction in other corners. ServiceNow, an IT service management and workflow automation company, this month said it intends to buy Swarm64 to expand the capabilities of running complex workloads on its Now Platform and to grow its presence in the open-source community. Key to Swarm64’s portfolio is providing a single database both operational and analytical data processes.

It makes sense for enterprises and smaller companies that are running MySQL workloads in the cloud, Nipun Agarwal, vice president of research and advanced development at Oracle, tells The Next Platform.

“We have MySQL customers and customers using other open source products,” Agarwal says. “Those cover a very, very wide range. Many of the large enterprises use them, but a lot of the very small enterprises also use them. This product is targeted towards all of those, so we are seeing a lot of the smaller customers because MySQL is very popular among them. Then you have a class of customers who are moving the data out from MySQL or other databases and using Redshift, Snowflake, Google Query, which are very, very expensive. Even those customers are saying, ‘Wait a second. Not only can I save money by HeatWave, not only can I get much faster by HeatWave, but I don’t even need to move data around and I don’t need to have data fragmentation. It’s a single database. The good thing is with this service is, we are seeing synergy and draw from both ends of the spectrum.”

Now Oracle is bringing greater automation and intelligence to HeatWave via machine learning techniques. The goal is to make it easier to use and to extend its performance and scalability, according to the Oracle officials. MySQL Autopilot, which is available now, automates such tasks as provisioning, data loading, query execution and failure handling. It also uses those machine learning techniques for jobs like sampling data, collecting statistics regarding data and queries, and building machine learning models using Oracle’s AutoML tool for modeling memory use, network loading and execution time.

“As a cloud provider and as a service provider, there’s a lot more opportunity we have to automate our customers’ experience,” Agarwal says. “In the cloud, we have a much better idea of the optimization targets customers have. Are they trying to optimize for throughput, for latency, for cost? We know exactly what configuration parameters have changed. We know exactly the stack of the software of the solutions – the database solution, the operating system solution. We know the underlying hardware and we can get real-time metrics. The approach we have taken for automation is to use machine learning-based automation, in contrast to rule-based automation. The advantage of machine learning-based automation, among other advantages, is that it provides a custom fit for the database. It’s not one-size-fits-all. The second thing is, as you run more workloads, as you run more queries on this database, the system learns from it and gets better.”

Oracle also is unveiling MySQL Scale-out Data Management. With this feature, HeatWave supports 64-node clusters – up from 24 node – and can process up to 32 TB of data. Previously the max amount was 12 TB. Agarwal notes that organizations are putting more data into their MySQL databases and the new scale-out tool improves the performance of reloading the cluster by up to 100 times.

“What we have now done is that when the user loads the data for the first time from the MySQL database into the cluster, the cluster also makes a copy of this in the media presentation on the objects store such that in the future, whenever there is a need to reload this data into the cluster, instead of going to the MySQL database, it can fetch this data from the object store,” he says.

Zivanic and Agarwal say that the scale-out capabilities improve the performance advantage against AWS. AWS remains the king of the public cloud space, holding onto 33 percent of a global cloud infrastructure services market that 39 percent year-over-year in the second quarter, hitting $42 billion, according to Synergy Research Group. Oracle still has a much smaller slice of the pie, though that means chipping away even a small portion of the market from AWS, Microsoft Azure or Google Cloud could mean big gains for the company.

Oracle also is keeping an eye out for smaller competitors. The vendor is now using benchmarks like TCPH to compare HeatWave on Oracle Cloud Infrastructure (OCI) to Snowflake, a fast-growing company that offers a cloud-based data platform. There the numbers show HeatWave being 6.8 times faster and 5.2 times cheaper.

“The challenge their users have is Snowflake basically excels in shoe sizes,” Zivanic says. “You can get 16 nodes, 32 nodes and so forth, so if you need 18 nodes, you have to buy 32, which makes their pricing astronomical. It’d be the equivalent of you and I going to Foot Locker and saying, ‘We need a size 10 shoe’ and they say, ‘Sorry, we have size 18’ and you and I walk around size 18 shoes. With Snowflake, what we anticipate there as customers look at these benchmarks and say, ‘Wait a minute. I get 7X faster performance at a fifth the cost and I’m not stuck just doing analytics. I can also run transactional, which Snowflake can’t do.’ It comes back to the whole business value analysis. If you want to look at the business value analysis, you start saying, ‘I don’t want to move data. ETL means surface-area exposures and security risks.’ Why would you want to do that?’”

Sign up to our Newsletter

Featuring highlights, analysis, and stories from the week directly from us to your inbox with nothing in between.
Subscribe now

Be the first to comment

Leave a Reply

Your email address will not be published.


This site uses Akismet to reduce spam. Learn how your comment data is processed.