Scaling Out Without Fault

Print Friendly, PDF & Email

Mike_AzevedoIn this special guest feature Mike Azevedo, CEO of Clustrix, discusses how significant spikes in simultaneous transaction processing are placing more demands on relational databases than ever. Mike is CEO of Clustrix, provider of the scale-out relational database designed to meet the needs of large, fast-growing e-commerce sites and other big Web- and cloud-based applications. He has more than 25 years of experience in scale-out analytic applications, grid computing, storage infrastructure, security, and retail, and has led top sales divisions at Platform Computing, ParAccel and PostX Corporation. Mike holds a Bachelor of Science degree in business administration/management from Chico State University

Relational databases always have been at the heart of online transaction processing (OLTP), and they will continue to be for the foreseeable future. But today the business world is demanding more out of this bedrock technology than ever. A variety of industries are experiencing massive increases in the amount of data that needs to be processed quickly and orderly, such as large-scale e-commerce sites juggling thousands of shoppers simultaneously generating cart activity. Or think of adtech companies that place millions of contextual ads in front of consumers each day. In these cases, databases are under pressure to handle high volumes and speeds without compromising a customer’s or end user’s experience. Fault tolerance and fast response times are critical. Today’s user who spends an inordinate amount of time on their mobile device is becoming more demanding than ever in terms of response time (i.e., latency) expectations. In many cases, a slow response is the same as none to someone accessing a site on their smartphone or tablet.

There’s an enormous demand for relational databases that can maintain performance (e.g., no site slowdowns or failures) while keeping the integrity of these transactions intact, whether that means seeing retail purchases through to completion or retaining the full record of vehicle and cargo tracking data over a period of time. Or to put it in more technical terms, organizations need their relational databases to maintain ACID (Atomicity, Consistency, Isolation, Durability) compliance while processing a deluge of transactions in such a tiny window of time.

Moreover, administrators want these relational databases to help organizations operate at the same brisk pace of business expected by consumers; they need on-demand, real-time reporting of up-to-date metrics—even the 10 to 30 minutes it takes to extract, transform and load a query from a separate analytics database won’t cut it anymore. Databases that can generate these reports at warp speed enable companies to adjust pop-up ad tactics, rearrange inventory, reroute fleets, troubleshoot anomalous sensor readings, locate missing data or otherwise course correct before your initiative or organization strays further off kilter.

A Crowded Marketplace

A few years ago, relational databases running in your data center were relatively straightforward to sum up; legacy relational databases like Oracle were proven for scaling out behind the firewall, while alternatives like MySQL effectively forced you to give up scalability in exchange for ease of use at a lower cost. Then a new generation of relational databases popped up to provide affordable scale, including the NewSQL databases like ClustrixDB, MemSQL and VoltDB.

Now, Amazon has vocally swiped at the establishment as it has advanced its Aurora relational database built in the Cloud, while other smaller players have also introduced cloud-only products.

Of course, most organizations don’t operate in a black-and-white world. Realistically, many have applications in the Cloud, in data centers and on premise, but cloud-only relational databases like Aurora can’t be used in the data center. Moreover, an RDBMS must be able to scale while handling a heterogenous mix reads and writes; Aurora touts benchmark results that operate in a vacuum and are not representative of reality. The data size used in its testing was small enough to fit on a cell phone, while workloads were run independently so as not to have reads conflict with writes—once again, not a real-world scenario.

At least for the foreseeable future, many IT professionals and their companies’ existing infrastructures need relational databases that can support a hybrid environment (i.e., data center and cloud), while continuing to deliver ACID compliance, low latency and high availability without even a slight fault in site performance. When your application is your business, you can’t afford downtime.

Demanding from Your Database What Your Customers Demand from You

Either way, companies in retail, adtech, IoT, travel and manufacturing, among other industries, are going to be under immense pressure to process a high volume of transactions in short order without a perceptible hiccup from an end user’s perspective. Databases of yesterday could do this in your data center at an enormous cost. But today there are a confluence of paradigms that enable businesses of all sizes to achieve this elasticity without breaking the bank or rearchitecting your IT infrastructure. The companies that will deliver the always-on availability, push-button simplicity, little-to-no latency and transaction reliability expected by today’s end user will find a way to combine the power and performance the old guard delivered on premise with the pay-for-only-what-you-use models that are likely to penetrate deeply into the corporate landscape over the coming years.

In today’s world, scalable, high-performance relational databases are available to everyone.


Sign up for the free insideAI News newsletter.


Speak Your Mind