Twenty years with MSPs

For more than twenty years, Kevin Hoffman and the team at eFolder, later Axcient, built backup and disaster recovery for managed service providers. The company was bootstrapped from nothing, reinvented several times, and grew into a market leader trusted by more than 5,000 MSPs before its acquisition by ConnectWise in 2024. Along the way, thousands of MSPs entrusted it with hundreds of petabytes of their clients' data, running on tens of thousands of servers and containers.

The daily fight

A fleet like that produces an ocean of logs and metrics, and that ocean was a daily fight. Open source observability tools were free the way a free boat is free: they demanded constant care from the SRE team and ran on expensive NVMe storage and compute in data centers. The alternative arrived by cold call, week after week: enterprise observability vendors with impressive demos and quotes that ran into millions of dollars a year, every year.

The breaking point

The breaking point was always the worst possible moment. Take Axcient's Microsoft 365 backup platform: when Microsoft's Graph APIs went down or throttled hard, the platform would throw off storms of HTTP 429 errors, and internal telemetry volume would spike 10x within minutes. Traditional observability clusters need minutes or hours to scale up.

The data layer

This motivated Kevin to go invent the data layer every SRE dreams of... Petabyte scale. Serverless, scaling from zero to ingesting petabytes in seconds. Nearly zero operational overhead: schemaless, with structure extracted automatically from raw text. Cost-effective enough that you never have to choose which logs to keep. It took years of intense engineering, battle-tested with early customers ingesting telemetry at serious scale. But we emerged with an o11y data layer that's truly a joy for SREs everywhere.

Then AI

Then AI changed what one engineer can do. AI agents are a force multiplier for human talent, but they are only as good as the data they can see. Solving a hard technical mystery takes evidence: structured, queryable, token-efficient evidence. The tools MSPs already use can't provide it, because they were never built on an observability data layer. Building one that handles petabytes and stays affordable is the hard part. It's the part we'd already done.

SparkLogs

That data layer is SparkLogs. AutoExtract turns unstructured text into structured data. Clustering and classification turn billions of messages into signal. Token-efficient query tools let AI agents gather cited evidence in seconds, at low cost. The result: AI with eyes across an MSP's entire fleet, and the hardest tickets closing in minutes instead of days.

Everything we ship obsessively carries our Spark promise: simple, powerful, fast.