Open, high‑performance infra alternatives arrive
New entrants are pitching open, high‑performance options for analytics: IOMETE launched a self‑hosted lakehouse on Kubernetes running Spark and Iceberg, aiming to avoid vendor lock‑in, and Arc Cloud highlighted a managed analytical DB built on DuckDB/Parquet with very high bare‑metal throughput. (x.com) Both plays emphasise open formats and deploy‑option choice rather than monolithic managed warehouses. (x.com)
Most companies still do analytics in one of two ways: they either rent a giant managed warehouse, or they glue together open-source parts and spend months babysitting them. Two new products are trying to make the second option feel much closer to the first. (iomete.com) (docs.basekick.net) A lakehouse is the idea behind one of them. It keeps data in cheap object storage as ordinary files, then adds table rules on top so the files behave more like a database than a folder full of comma-separated values. (iomete.com) (iceberg.apache.org) IOMETE is selling that model as a self-hosted stack you run on Kubernetes, which is the software many companies already use to schedule containers across clusters. Its platform is built around Apache Spark for compute, Apache Iceberg for table metadata, and object stores like Amazon Simple Storage Service, Google Cloud Storage, Azure Blob, MinIO, and Dell Elastic Cloud Storage for the data itself. (iomete.com) (spark.apache.org) Apache Spark is the part that does the heavy lifting. It spreads jobs across many machines, so a query over terabytes of logs works more like a team carrying a couch than one person dragging it alone. (spark.apache.org) Apache Iceberg is the part that keeps those files organized. It tracks snapshots, schema changes, and partitions in metadata, so users can treat piles of Parquet, ORC, or Avro files like a consistent table instead of manually guessing which files belong together. (iomete.com) (iceberg.apache.org) That combination is aimed at a specific buyer: a team that wants cloud-style analytics without handing over its storage format or deployment model to one vendor. IOMETE says it ships as a Helm chart into a Kubernetes cluster and gives one console for Spark jobs, SQL queries, catalog browsing, and security policies. (iomete.com) Arc is coming at the same problem from the other side. Instead of a full lakehouse control plane, Arc pitches a managed analytical database whose core engine stores data in Apache Parquet files and runs queries with DuckDB, the in-process analytics engine that has become popular for fast local analysis. (basekick.net) (docs.basekick.net) DuckDB is built for scanning columns very quickly, which is why it shows up in so many notebook and embedded analytics workflows. Arc wraps that engine in a database product and claims more than 18 million records per second of ingestion and more than 6 million rows per second of queries. (docs.basekick.net) (github.com) Parquet is the storage piece that makes Arc’s pitch legible to data teams. Because Parquet is an open columnar file format, the data is not trapped in a proprietary layout, and Arc says those files can live on local disks or cloud object stores such as Amazon Simple Storage Service and Azure. (docs.basekick.net) The split between these two products is really about where they put complexity. IOMETE assumes you want a full open lakehouse you can deploy inside your own infrastructure, while Arc assumes you want a simpler database surface with managed service convenience on top of open files and a fast query engine. (iomete.com) (docs.basekick.net) That is why both launches feel like part of the same shift. The sales pitch is no longer “come into our warehouse and stay there forever”; it is “keep your data in open formats, pick where the software runs, and still get performance numbers that sound like a modern warehouse.” (iomete.com) (basekick.net)
