ClickHouse vs MongoDB

Introducing ClickHouse's new TimeSeries Engine: Your drop-In Prometheus replacement

ClickHouse PromQL support lets you store Prometheus metrics in ClickHouse Cloud, query them using familiar PromQL, and bring metrics together with your logs and traces without rewriting queries in SQL.

High-Performance ETL for Apple Health XML Exports

This technical tutorial describes building a high-concurrency ETL engine to process very large Apple Health XML exports. The author outlines a performance-first architecture: a streaming Rust XML parser (quick-xml) to extract <Record> tags, export data as Apache Arrow record batches via PyO3 for zero-copy consumption in Python/Polars, and bulk-ingest cleaned data into ClickHouse using clickhouse-connect. The post includes code snippets for the Rust parser, Arrow-to-Polars bridge, ClickHouse table schema and bulk insert, and discusses production considerations such as parallelization, schema evolution, malformed XML handling, Grafana visualization, and feeding Arrow buffers into ML frameworks like PyTorch.

• Apple Health exports a single massive XML file containing millions of <Record> tags.
• The pipeline uses a streaming Rust parser (quick-xml) to avoid loading the entire XML into memory.
• Rust code is wrapped with PyO3 to return Apache Arrow tables for zero-copy consumption by Python/Polars.

NeverBlink Adds ClickHouse Support

NeverBlink, an AI-driven database administrator, announced support for ClickHouse, extending its database resilience and optimization platform to ClickHouse clusters. The integration offers continuous health checks, automated root-cause analysis, query analytics, prioritized remediation guidance, and optional unlimited expert consulting via BigData Boutique. NeverBlink already supports Elasticsearch and OpenSearch and is used in production by companies including Akamai, Perforce, and Palo Alto Networks. The article notes ClickHouse has surpassed 4,000 customers and a reported $15 billion valuation. Leadership quotes emphasize reducing downtime, lowering costs, and accelerating database optimization in the AI era.

• NeverBlink announced support for ClickHouse, extending its AI-native database resilience and optimization platform to ClickHouse clusters.
• NeverBlink is described as the world’s first AI database administrator (DBA) and already supports Elasticsearch and OpenSearch.
• ClickHouse has reportedly surpassed 4,000 customers and a reported $15 billion valuation.

Bank of America Backs MongoDB, Raises Price Target

Bank of America reiterated its buy rating on MongoDB and raised its 12-month price target to $540 from $450, citing accelerating AI adoption that boosts demand for data management systems. Analyst Koji Ikeda said MongoDB is positioned to win meaningful share of future AI workloads, praising its ability to handle large datasets, memory, scale, and real-time transactional data. The bank urged investors to view any fear-driven weakness as a buying opportunity. FactSet and LSEG data cited in the piece show MDB has rebounded strongly year-to-date and that the majority of Wall Street analysts rate the stock buy or strong buy.

• Bank of America reiterated its buy rating on MongoDB and raised its 12-month price target to $540 from $450.
• Bank of America analyst Koji Ikeda wrote that "MongoDB will win [a] meaningful share of future AI workloads."
• MongoDB was said to have rebounded 86% since March 31 and has more than doubled in the past year, per FactSet data.

Decoupling CRM from MDM for Device Management

The article describes an architecture for integrating Mobile Device Management (MDM) with an internal CRM without making the MDM a core dependency. Using NestJS, MongoDB/Mongoose, and TanStack Start with Fleet as the initial provider, the author separates business state (CRM-owned device records, device assignments, and immutable device action audit logs) from technical state (MDM-owned OS, hardware IDs, last check-in). Key patterns include a DeviceProvider abstraction (so different MDMs like Fleet, Intune, or Jamf can be swapped), a background DeviceSyncWorker that synchronizes technical device data into MongoDB every five minutes, and creation of DeviceAction audit records before executing destructive operations (lock/wipe) via the provider. The result is vendor independence, faster reads, auditable operations, and simpler frontend development.

• Author implemented an MDM-agnostic Device Management module using NestJS, MongoDB/Mongoose, and TanStack Start with Fleet as the initial MDM provider.
• The CRM owns three business concepts: Device (with a decoupled providerId), DeviceAssignment (historical ledger), and DeviceAction (immutable administrative audit record).
• A DeviceProvider interface defines required provider operations (getDevice, listDevices, lockDevice, wipeDevice); provider-specific implementations (e.g., FleetProvider, IntuneProvider) satisfy that interface.

How to Export FTDC From MongoDB Atlas

This technical how-to shows how to extract FTDC (Full Time Diagnostic Data Capture) from MongoDB Atlas using an undocumented Atlas Admin API (v1.0) that creates a log collection job, polls for success, and downloads a bundle containing per-member diagnostic.data directories including the live metrics.interim file. The author lists the three required inputs (programmatic API key, project ID, replica set name), provides example curl calls, and describes bundle metadata (expiration ~30 days). Atlas sets diagnosticDataCollectionDirectorySizeMB to 400MB, which typically yields 2–5 days of FTDC under real load (longer on idle clusters). The post also warns that the built-in clusterMonitor role grants read access to the oplog and demonstrates a more restrictive diagnostics-only custom role. The author references tooling (Big Hole, keyhole) for reading FTDC locally.

• MongoDB servers write FTDC (Full Time Diagnostic Data Capture) into a diagnostic.data folder with roughly 5,700 metrics sampled once per second.
• MongoDB Atlas exposes a v1.0 Admin API endpoint (logCollectionJobs) that can package FTDC on demand, requiring a programmatic API key, project ID, and replica set name.
• The FTDC extraction flow uses three API steps: create a logCollectionJob, poll the job until SUCCESS, and download the resulting bundle (example curl sequence provided).