October 2026 Summaries
10 posts from Axiom
Filter
Month:
Year:
Post Summaries
Back to Blog
Axiom announced its status as a silver sponsor of Let’s Encrypt, the free and automated certificate authority focused on expanding HTTPS adoption. The company says Let’s Encrypt enables it to automatically provide every Watchly instance with a free security certificate, protecting customers without requiring user intervention. Axiom supports Let’s Encrypt’s mission of securing machine-to-machine communication and views encrypted connections as essential both across the internet and within internal infrastructure. Let’s Encrypt Executive Director Josh Aas welcomed Axiom’s support and its embrace of automation and ease of use for HTTPS deployments, while Axiom encouraged website operators to adopt HTTPS using the organization’s compatible libraries and community resources.
Oct 29, 2026
251 words in the original blog post.
APL’s new `sample n` operator returns an exact number of rows selected uniformly from all query matches, addressing the bias of combining probability-based filter sampling with an early `take` limit, which tends to favor data scanned first. It uses reservoir sampling, specifically Algorithm R, to process streaming data with constant memory while giving every matching row an equal eventual chance of selection. In distributed queries, each worker maintains a reservoir and reports its total number of matched rows, while a weighted coordinator merge uses hypergeometric allocation to determine how many sampled rows each worker should contribute according to its share of the overall population. This avoids the bias of generic merges that treat equal-sized worker reservoirs as representing equal amounts of data despite differing match counts. The implementation required a dedicated distributed sampling plan and makes count propagation part of the operator contract, although it currently applies only to single datasets and still scans every matching row; future work aims to support unions and reduce scanning without sacrificing fairness.
Oct 28, 2026
2,258 words in the original blog post.
Axiom Cloud has introduced query budgets that let organizations set hourly and daily dollar-based limits on data queries for API tokens and user roles, addressing the increased volume of observability investigations performed by AI agents. Agents using Axiom’s MCP Server operate under an engineer’s permissions and personal role budget, while autonomous services such as scheduled jobs, CI scripts, and AI teammates can receive dedicated tokens with restricted permissions and independent spending caps. Budgets limit only query activity rather than data ingestion, ensuring agents can access the same complete data as human users while preventing any person or token from exceeding its allocated query cost. Organizations can apply budgets across built-in roles, API tokens, and, with the RBAC add-on, custom roles and groups; when users hold multiple roles, the smallest applicable budget is enforced. Agents receive query cost and remaining-budget information, and exhausted budgets stop only the affected user or token until their hourly or daily UTC reset, while dashboards, monitors, and other users continue functioning. These controls complement, rather than replace, an organization-wide monthly spending limit that covers both query and ingestion costs.
Oct 21, 2026
889 words in the original blog post.
Axiom rewrote the parsing stack for its open-source Metrics Processing Language (MPL), replacing the pest parser generator with a hand-built system after discovering that pest’s rapid-development advantages did not extend easily to features such as resilient error recovery, syntax highlighting, and code formatting. Rather than framing the change as a simple technological upgrade, the account emphasizes that pest helped the team learn MPL’s requirements and identify the limitations of maintaining separate lexer, parser, and formatter implementations. The new design uses a stateful lexer capable of handling nested string interpolation, Rust Analyzer’s rowan crate to construct a lossless syntax tree that preserves every byte of source input including whitespace and comments, and a final transformation into an abstract syntax tree for execution. Each stage handles a distinct category of errors—invalid lexemes, syntactic problems, and semantic rule violations—allowing later stages to rely on guarantees from earlier ones while supporting recovery beyond the first error. Although error recovery still needs refinement, the architecture is intended to make it easier to improve, and an interactive WebAssembly playground demonstrates the real MPL lexer, parser, and interpreter in a browser.
Oct 20, 2026
1,426 words in the original blog post.
Observability spending cannot be meaningfully reduced to a universal percentage of an IT budget because requirements vary by architecture, compliance, telemetry volume, retention needs, and investigation practices; instead, the central measure is whether teams retain complete, searchable evidence long enough to investigate incidents. Costs may grow independently of traffic when vendors charge by hosts, metric-series cardinality, indexed-log volume, or multiple pricing dimensions, particularly in microservice, Kubernetes, and OpenTelemetry environments with many attributes and short-lived workloads. The discussion cautions that sampling, limited retention, and separate indexing tiers can reduce the usefulness of collected data, even when a bill appears affordable, and recommends assessing what proportion of loaded data remains queryable and for how long. It cites Gartner figures indicating that 36% of surveyed clients spend over $1 million annually on observability and that logs account for more than half of spending, while arguing that such figures do not reveal the quality or accessibility of retained data. Axiom presents its alternative pricing model as usage-based charges for data loading, query compute, and compressed storage, with no cardinality or separate indexing charge, automatic volume discounts, free-tier options for smaller teams, and optional governance features, while positioning its costs as lower than several named competitors under selected workload comparisons.
Oct 17, 2026
2,177 words in the original blog post.
Splunk’s Machine Data Lake stores raw machine data in managed object storage and supports limited raw searches, while operational capabilities such as dashboards, alerts, scheduled searches, correlation, and broader event search generally require data to be promoted into a Splunk index. This architecture separates costs for raw data landing and retention, promotion processing, and storage of promoted copies, with static promotion serving historical data ranges and streaming promotion handling future events without backfilling history. Eligible Splunk Cloud customers may receive included Cisco Data Fabric capacity, but the text states that complete public per-GB pricing for additional landing, index promotion, streaming promotion, and retained copies is not available, making detailed quotes necessary for cost comparisons. It contrasts this approach with Axiom’s published usage-based pricing for data loading, query compute, and compressed storage, which it says keeps loaded data queryable without a separate promotion step or duplicate retained copy. The piece also presents Axiom as capable of coexisting with Splunk through HEC-compatible ingestion and federated-search integrations, enabling organizations to compare both platforms using the same workloads.
Oct 17, 2026
2,300 words in the original blog post.
Splunk’s core platform is an index-centric search and analytics system rather than a traditional data lake, because it parses and indexes events at ingestion to support fast SPL searches, dashboards, alerts, and correlation workflows. Its newer Machine Data Lake adds a separate raw-data retention tier in managed object storage, where supported logs and events can be stored and searched with limited fields, but data generally must be promoted or dual-routed into a Splunk index before it can support standard operational dashboards, alerts, and scheduled searches. SmartStore differs from both by moving already indexed warm buckets to remote object storage, although uncached data may need to be downloaded before searches run. The text contrasts this layered model with Axiom’s architecture, which it describes as retaining events in a single query path from recent live data through compressed object-storage blocks, without a separate promotion process. It also notes that organizations evaluating a workload should determine whether data resides in Splunk’s index, SmartStore, the Machine Data Lake, or both, since that location affects search speed, monitoring capabilities, retention practices, and integration options.
Oct 16, 2026
1,741 words in the original blog post.
Axiom promotes a startup program for Seed through Series A companies that are new to paid Axiom Cloud and apply through eligible accelerators, investors, or startup communities. Approved participants receive a waiver of the $25 monthly platform fee for 12 months and a 30% discount on Data Loading and Query Compute, capped at $10,000 in compute savings, while retaining monthly included usage of 1 TB of data loading and 100 GB-hours of query compute. Storage and add-ons remain subject to standard pricing, and the compute discount cannot be combined with committed-use or other promotional compute discounts. The page also illustrates Axiom’s log-query interface, showing filters for HTTP error statuses and U.S. traffic alongside detailed request and Vercel deployment metadata, and includes a customer endorsement emphasizing the platform’s usability.
Oct 08, 2026
1,128 words in the original blog post.
A security data lake is described as a high-volume, full-fidelity, continuously queryable repository for security telemetry that complements rather than replaces a SIEM. While SIEM platforms remain responsible for detection rules, alerting, case management, and response workflows, the data lake stores raw and normalized events for longer periods at lower cost, avoiding sampling or short retention that can obscure low-and-slow attacks. The recommended approach includes ingesting sources such as cloud audit logs, VPC flow data, WAF events, endpoint telemetry, and identity logs; normalizing commonly used fields while retaining raw payloads; using an event store that supports affordable compressed retention without rehydration; and providing an ad hoc investigation-oriented query interface. Retention should be based on compliance and investigative requirements, with access controls, auditing, encryption, and tenant isolation supporting governance. Deployment can begin alongside an existing SIEM by sending high-volume or previously unaffordable sources to the lake while leaving SIEM detections and cases unchanged. The source presents Axiom as an example of an event-store and query-layer provider, offering schema-less ingestion, virtual fields, configurable retention, and its APL query language, while explicitly leaving alert correlation and case management to the SIEM.
Oct 02, 2026
1,889 words in the original blog post.
Security data lakes are presented as a cost-effective way to retain large volumes of security telemetry, such as cloud audit trails, network flow logs, endpoint events, and identity activity, while SIEM platforms continue to handle detections, correlation, and case management. The central distinction is whether retained data remains directly searchable or must be restored, promoted, or reindexed before analysts can investigate historical events; the latter approach can introduce delays and encourage organizations to sample or discard high-volume logs, potentially creating investigative blind spots. The discussion recommends that buyers assess a vendor’s ingest-to-query process, historical query performance, filtering practices, automated-agent access, and security controls including RBAC, SSO, audit logging, encryption, and tenant isolation. Axiom positions its platform as an alternative architecture in which immutable, compressed event data remains in a unified query path across both recent and long-retained telemetry, allowing SIEM workflows to remain in place while extending searchable retention.
Oct 02, 2026
1,920 words in the original blog post.