Data observability startup Acceldata Inc. today announced that it has raised $8.5 million in a Series A venture capital round, bringing its total funding to $10.5 million.

The company, which was founded by four former Hortonworks Inc. engineers, is targeting machine learning and analytics teams with a platform that can find and resolve problems with data.

Observability is a growing category of instrumentation software that goes beyond flagging errors to diagnosing root causes. Most observability tools target infrastructure, applications or websites but Acceldata is the first to focus on data, said Chief Executive Rohit Choudhary.

“Observability requirements for analytics and artificial intelligence are very different from those of web apps,” he said. “We are creating a new market for ourselves.”

Acceldata said its platform synthesizes signals across workloads, infrastructure and usage into a single view that enables customers to predict, identify and fix data issues. Few but the largest technology companies have the experience and resources to build observability tools that work across distributed data systems at scale, Choudhary said.

“If you’re trying to derive those insights you find that tools at your disposal are pretty rudimentary,” he said. “We want to democratize highly scalable observability.”

Among the factors he said are driving the need for data observability are the shift from monolithic applications to microservices and the often overwhelming amounts of data that are required to train machine learning algorithms. With the shift to microservices, “businesses are no longer as dependent on applications. They’re dependent on data directly,” he said.

Data errors can cause analytics and machine learning algorithms to go off track and deliver misleading results. Engineers running analytics frameworks like Apache Spark “have to run across thousands of boxes to find the right data,” he said. “You end up spending time digging through server logs to find out what failed.”

Acceldata works with open-source technologies such as Apache NiFi, Druid, Spark, Spark Streaming, Kafka, Flink, HBase and Hive, as well as cloud services that include Amazon Web Services Inc.’s EMR and Google LLC’s Dataproc in any combination. The company said its platform recommends ideal configurations and provides a unified command center for multi-cluster data and workload management. It can also predict when capacity thresholds are about to be exceeded.

“We’re trying to reduce the amount of data that comes in and simplify for the users who are trying to optimize,” Choudhary said. Installation across a 500-node cluster takes just 45 minutes to “generate a complete map of the system and deliver insights the next day,” he said. Among the elements discovered are physical layout, running services, where inefficiencies exist what activities users are performing.

The new funding was led by new investor Sorenson Ventures Associates LP with participation from previous investor Lightspeed Venture Partners and Emergent Ventures LLC.

Image: Marco Verch/Flickr CC