The need for faster, more scalable distributed databases has spawned a host of in-memory cluster-oriented database offerings. But while raw speed and scale hold sway in the early going, the innovators will eventually move to add management capabilities for traditional DBAs, as a recent MemSQL product update indicated.
Earlier this year, in-memory database provider MemSQL Inc. released MemSQL Ops dashboards intended to allow database administrators (DBAs) to continuously monitor cluster resources, and even pinpoint the most active tables in any cluster, which will better enable DBAs to plan cluster growth and diagnose performance issues.
"There is a real difficulty in working with distributed systems when you are just using a command-line interface," said Eric Frenkiel, co-founder and CEO of MemSQL, explaining the rationale behind the company's Ops dashboards. The product enhancements are meant to complement advanced takes on data processing that Frenkiel and co-founder and CTO Nikita Shamgunov devised after leaving Facebook, where fast, high-volume analytics were a goal.
"We wanted to connect a lot of commodity machines together, but to provide a standard ANSI SQL database at the same time," Frenkiel said. What he and Shamgunov devised was a shared-nothing data platform combining an in-memory row store with a flash, solid-state drive (SSD) or disk-based compressed column store.
The system can perform in both real time and batch applications, said Frenkiel, who noted that fast analytics has become a major use case for MemSQL recently.
"MemSQL seems to have a really solid, if somewhat focused, product," said Nik Rouda, analyst at Enterprise Strategy Group. "It's an extremely high-performance system, and looks quite easy to use," he said, indicating the recent operations management dashboards contribute to that ease of use.
Meanwhile, forays into in-memory technology like MemSQL's are likely to be repeated across database offerings, according to Rouda. "Memory is becoming one of the hot things in the space -- there are fundamental physics advantages," he said, also pointing to a multiyear trend in reduced semiconductor memory costs as a driver.
SyncSort set to transform Amazon cloud data
It's not just for social media and sales leads anymore -- the cloud may someday be the primary residence for traditional enterprise data. That could even include determinedly down-to-earth mainframe workloads that few people expected would ever take the cloud route.
Extract, transform and load (ETL) specialist SyncSort is looking to spur this transition by providing ETL software for users moving mainframe data -- as well as legacy data warehouses and other jobs -- to popular Amazon cloud offerings. A recent software release from SyncSort provides graphically based ETL programming to data managers that want to combine the old (mainframe) with the new (cloud).
If the trend continues it could create a new balance of power. When generating reports, it may be increasingly necessary to match cloud and on-premises data.
"What we are seeing today is that a lot of data is originally gathered in the cloud. And, at the same time, the types of processing done on-premises can now be done more economically in the cloud," said Gary Survis, CMO at Syncsort. And some organizations are looking into moving the on-premises material up, rather than bringing the cloud data down, and such capabilities are now supported by the SyncSort line, Survis said.
As a follow-up to the company's release last year of IronCluster ETL for Hadoop, the recent cloud release is dubbed Ironcluster ETL on Amazon EC2. Once on EC2, data can be loaded to Amazon Redshift for analytics and reporting, Survis said. The data can also tap into Amazon Simple Storage Service for log data for aggregation, and connect to Amazon's high-throughput DynamoDB NoSQL database.
Find out more about new approaches to SQL systems
Learn about key trends in managing big data environments
Check out Hadoop's role in mainframe modernization