一位具有Meta背景的技术人士批评OpenTelemetry将可观测性割裂为"指标、日志、追踪"三个部分，认为这种分类方式造成了行业混淆1。该人士指出，OpenTelemetry官方术语表已包含超过60个术语，进一步加剧了业界理解的难度1。
该人士主张采用Meta的Scuba系统所使用的"宽事件"（Wide Events）概念作为统一的可观测性基础1。宽事件以JSON文档的形式表现，包含字段名和对应的值1。在这一模型下，指标、日志和追踪都可被视为宽事件的特殊情况，而非相互独立的概念1。Scuba系统内置原生采样机制，能够自动调整结果而无需手动计算1。
该人士认为，宽事件模型提供了更强的探索能力和处理未知问题的潜力1。与此同时，Honeycomb和Axiom等厂商正在提供更为简洁自然的可观测性交互方式1。
An author drawing on experience at Meta has challenged the prevailing observability framework promoted by OpenTelemetry, arguing that its division of observability into three distinct categories—metrics, logs, and traces—creates unnecessary confusion across the industry. 1 The author contends that metrics, logs, and traces are merely special cases of a more fundamental concept: Wide Events, a unified data structure that Meta developed within its Scuba system. 1
According to the argument, Wide Events represent a simpler and more powerful foundation for observability than the fragmented three-pillar approach. 1 The Scuba system implements Wide Events as JSON documents containing field names and values, and features native sampling mechanisms that automatically "scale up" results without requiring manual calculation. 1 The author makes the case that Span—a core concept in tracing—is effectively a Wide Events structure, and that both logs and metrics can be mapped onto the Wide Events model. 1
The critique also points to the complexity within the observability ecosystem itself: the OpenTelemetry website's terminology glossary contains over 60 terms, contributing to industry-wide confusion. 1 Meanwhile, observability vendors including Honeycomb and Axiom have begun offering simpler and more intuitive approaches to observability interactions, suggesting that broader adoption of a unified Wide Events model could align with market trends toward usability and clarity. 1
评论
还没有评论，欢迎留下第一条。