Marty Kagan and the bill that wouldn't go away

The bill was becoming difficult to ignore. At Cedexis, Marty Kagan and his colleagues were collecting the records that explained how the internet was behaving. More customers meant more records. More records meant more money spent keeping them. Eventually, storing the logs essential to the business ranked among its largest expenses, behind payroll. Success had acquired a rather expensive filing cabinet.

Kagan had helped build a company that made content delivery easier to understand. Now understanding it was generating a second problem. Throw records away and you lost detail. Keep them and the expense grew. Accept slower answers and the usefulness of the whole exercise suffered. The choices were familiar enough to seem permanent.

Hydrolix, which he co-founded with Hasan Alayli in 2018, grew out of their refusal to leave that arrangement alone. Its origins are unusually concrete: two colleagues remembered a cost they had actually paid. Kagan's subsequent work has involved both rebuilding the machinery underneath those costs and finding a practical way to put it in customers' hands.

A history student gets the web assignment

Before the databases, there was a history department. In 1992, Kagan completed an undergraduate thesis at UC Santa Barbara titled Monkey in the Middle: Lithuania's Role in Nazi-Soviet Relations. His academic path also took him to UC Irvine. He earned undergraduate and master's degrees in California, then left a PhD program for a startup job in Santa Barbara.

A small item in UCSB's October 1996 history newsletter catches him in that transition. It credits former student Marty Kagan, then working at the Web Power Authority, with designing the department's initial homepage. Graduate students subsequently expanded the structure he had established that spring. Faculty profiles, research, syllabuses and announcements were finding a place online.

The newsletter recommended Netscape 3.0 and Internet Explorer 3.0. It admired the site's accessibility on newer Pentium computers. The details date the moment more effectively than a photograph of an old monitor could. The web was still something a university department introduced to its readers with directions.

It is an appealing early entry in a career later devoted to internet infrastructure: a history student helping make other people's work available through a browser. Years later, the records would be arriving billions at a time. The department website, at least, did not require a Super Bowl operations team.

The internet, viewed from the delivery entrance

Kagan went on to Cisco and Akamai, helping develop Akamai's West Coast research and development operation. His career also included engineering leadership at Jive Software and Fastly. These were jobs close to the practical business of moving information and keeping services working, rather than merely describing what the internet might someday become.

At Akamai, he also served as director of technology, international, based in Paris from 2004 to 2007. That chapter adds a geographical loop to the story. A company whose infrastructure career included a stretch in France would later grow through partnerships serving customers well beyond its Portland base.

Cedexis made the delivery problem measurable. Its Radar product assessed the experience content delivery networks provided, helping customers choose which network to use in different countries and make routing decisions. Alayli joined the engineering team in 2012 after working at Cisco. The logs were part of how the company understood performance.

The pair eventually took separate jobs, Alayli at Splunk and Kagan at Fastly. Citrix acquired Cedexis in early 2018. Yet the storage problem survived those changes of employer and ownership. When Alayli decided to pursue a different approach to the underlying technology, he called Kagan. They already shared the necessary introduction: both knew the bill.

Keep the records. Change the machinery.

In a January 2021 account of Hydrolix's first commercial offering, Kagan put the earlier company's volume at more than 15 billion transaction logs a day. At that scale, the unglamorous decisions become consequential. How is information indexed? Where does it live? How much computing capacity must remain attached to it?

Hydrolix separated storage from computing and built around inexpensive object storage, with compression and indexing intended to keep queries responsive. The phrase the company used was a streaming data lake: incoming records could become useful quickly while remaining available for later investigation.

Consider the everyday question beneath the technical vocabulary. An operations team sees trouble now. It wants recent details immediately, then wants to compare them with an older event. Moving historical records into an archive that takes time to restore makes that second question harder. Kagan's proposition was to make keeping and questioning the record less financially punishing.

He has also described boundaries. In his April 2025 conversation with Alan Shimel, he emphasized immutable, structured, time-series logs as Hydrolix's focus. It was a deliberately chosen workload. That specificity makes the business easier to understand: a company built around a recurring problem its founders had encountered, with engineering choices shaped by that problem.

“We started Hydrolix to change the economics of big data.”

Marty Kagan · January 2021

The second engineering problem was buying

A working database still has to get through purchasing. Kagan's account of Hydrolix's sales process is refreshingly numerical. Direct deals could take nine months and convert at about 30 percent. With partners that already had customer relationships, he said, the average cycle shortened to 60 days and conversion reached 70 percent.

Those figures describe his experience, rather than a promise about every future deal. They explain why partnerships occupy such a large place in this story. Customers were already buying services from established providers. Hydrolix could supply the data layer behind an offering those providers brought to market.

Akamai's TrafficPeak put that approach into practice. It runs on Akamai Connected Cloud using Hydrolix technology, giving customers a managed way to ingest, retain and investigate their logs. For Kagan, whose career had already included Akamai, an earlier workplace became a route to customers.

By November 2023, he was describing the wider opportunity for other observability providers: retain their interfaces and integrations while replacing the expensive data layer underneath. It is a business proposition with a certain courtesy. Customers and partners have already invested in tools; a new infrastructure company can make itself useful within those existing arrangements.

A football game with 55 billion records

During FOX's stream of the February 2025 Super Bowl, the proposition faced a very public sort of examination. Hydrolix handled nearly 200 terabytes of log data and 55 billion records. The operation brought together logs from four content delivery networks, arriving through different routes, for a consolidated view of delivery.

The numbers distinguish two kinds of speed. Newly arriving events became visible for analysis in five to ten seconds. Median query response time was 0.481 seconds. One measures the journey into the system; the other measures the answer after someone asks a question. Confusing them would make for a tidier boast and a less useful explanation.

Behind those figures were large Kubernetes deployments the team nicknamed Kaiju, after the Japanese movie monsters. Once the event finished, stateless computing let the team scale down while retaining the records in object storage. A championship game requires a temporary crowd. Its computing bill need not insist that everybody stay.

Growth brings a different kind of workload

In April 2025, Hydrolix announced an $80 million Series C led by QED Investors. Kagan reported eightfold revenue growth in 2024. He used the funding announcement to discuss integrations, partnerships and hiring, including connectors for systems such as Splunk and Databricks. The work was expanding from proving the engine to making it fit more environments.

By March 2026, Hydrolix reported more than 650 customers, compared with four two years earlier, and more than 200 employees, up from 40. It placed fourth on Inc.'s 2026 Pacific regional list. Those are company milestones, but they also change the daily job of its co-founder: more people, more customers, more decisions that cannot all run through one desk.

In May 2026, the company appointed Brian Howie chief revenue officer and Enrico Risi vice president of global strategic sales. In September, its IBC plans included partner demonstrations involving agentic AI and multi-CDN operations. The partnerships were still doing practical work, connecting the underlying platform to specialized applications and established industry relationships.

A calmer founder, with things to do outside

Kagan describes himself as calmer the second time around. His interests include gardening, cooking, reading and scuba diving. His advice to other founders begins with a plain instruction: “First, make sure you clearly understand what problem you're solving.” There is enough in his own career to explain why he returns to it.

The company's stated values give that managerial outlook a public expression: family first, leading by example, clearing obstacles and sharing information. They sit alongside an instruction to build around real needs and validate with evidence. For a business made from records of what actually happened, that last preference seems particularly appropriate.

At Fastly, one of his direct reports credited him with expanding a team that needed to handle company growth and engineering debt while keeping its core service running. The account describes clear priorities, authority over technical decisions and responsibility for the results. It gives the management language some practical detail. Hiring capable people only solves part of the problem; those people also need enough room to make decisions. The same colleague said the organizational practices Kagan introduced continued after he left.

There is a pleasing continuity here. A former history student helped put a department's work on the web, spent years inside the systems delivering it, then helped build a company concerned with preserving the records those systems produce. The scale has changed beyond recognition. The usefulness of an accessible record has held up rather well.