25 years of JOX, from Talkline to freenet.

1995 – 2000: From script chaos to a custom tool

The JOX story begins in 1995 at Talkline GmbH, then a young mobile operator in the growing German market. Day-to-day operation of the billing system at the time was handled by a large body of shell scripts with embedded sqlplus calls. These scripts were started and monitored manually. Errors had to be detected and fixed individually; visibility into the overall progress of the daily runs was barely there.

From this environment, a first tool gradually emerged at Talkline, developed by Alexander Odesser in close coordination with the system and billing analysts in Talkline's service office: a Perl-based server application accessing an Oracle database, with a GUI in SQL-Windows. It was called "Job Maintenance". It worked very well – but only on a single server. It wasn't designed for the growing demands of a multi-server landscape.

In the following years, the same author built a second generation at a successor company: the "Job Manager" – multi-server-capable, cleaner – but it was never rolled out at Talkline. Job Maintenance simply ran too well there to be replaced.

The trigger came in 2000: Talkline decided to replace its aging billing system with a new platform from an international vendor. The promise included a modern scheduling layer that was supposed to carry batch operations. In practice, the bundled layer turned out not to be sufficient for Talkline's real-world operation. Routine flows had to be run mostly by hand – with all the consequences for error rates, night shifts and visibility.

At this point, Talkline called the developer of Job Maintenance. Now a freelancer, he took on the engagement to evolve the internal tool to the point where it would carry the new operation. JOX was created from exactly that engagement.

2001: Greenfield with experience

JOX was built from scratch in Java – multi-server-capable from day one, with a central repository and distributed server processes. A migration function read job definitions directly from Job Maintenance, so that the Talkline know-how accumulated over years was not lost.

Rollout took five to six months. The approach was iterative and closely tied to the actual migration needs. The developer's familiarity with the industry — and with Talkline itself — kept the coordination loops with the service team short.

Process automation for the new billing system and the JOX rollout ran in parallel, step by step. Instead of pre-defining a fixed end state, requirements were continuously prioritized: what worked went into production; what became important next was pulled in.

The Talkline colleagues actively shaped both sides: the processes to be modelled in JOX, and JOX itself – through requirements, tests and fast feedback from production. It was exactly that day-to-day interaction that turned a tool for a single customer into a platform that has benefited other deployments ever since.

From crisis mode to steady operation

The new billing system went into production; the processes had to run 24×7. The day-to-day operation at Talkline involved around 20 to 30 in-house staff plus 10 to 15 external consultants from the platform vendor. Three to four students additionally monitored the end-of-day processing at night. Errors were frequent during this phase – newly introduced billing logic meeting newly introduced interfaces is normal start-up pain, and without a load-bearing scheduling layer, each individual error multiplied into follow-up work.

With JOX in place, the team gradually took over the triggering, monitoring and handling of these processes. The analysts modelled the dependencies between processing steps as JOX groups/workflows, defined error handling and retry logic once, and let them run in JOX. What previously needed human attention now ran deterministically. The actual errors – the substantive ones, in the billing logic itself – only became visible once the noise of the manual mishaps had disappeared.

Three concrete shifts made the effect of the rollout visible:

• Three to four students at night were no longer needed for end-of-day processing. The run coordinated itself; escalations only came up in exceptional cases.
• External consultants from the platform vendor were needed less and less for day-to-day operation. The Talkline team could carry the operation themselves.
• The error rate dropped to individual, substantively grounded cases. The error class "script didn't start / script crashed / nobody noticed" disappeared.

Within a few years, the JOX installation at Talkline grew to around 1,000 jobs per day. Spread across three to five Solaris servers, three to five Oracle databases and three to four areas: one cross-market plus market-specific ones for the individual markets. A typical daily distribution comprised 50 to 100 jobs for a billrun (which itself ran roughly a day), 100 to 200 for end-of-day processing, plus around 100 for other processes.

The Talkline team's work changed fundamentally. Instead of starting scripts manually and hunting for errors, the colleagues modelled new processes and brought them into JOX. JOX was intended as a platform from the start, not as a workaround for the time of migration. In the following years that's exactly what happened: JOX became the central control for every new billing process at Talkline.

Acquisition and comeback

In the years after Talkline's peak phase, the company became part of a larger mobile operator. The customer base was migrated into the in-house billing system. And the in-house scheduler came with it.

In the new operation, that scheduler didn't prove itself to the same degree that JOX previously had at Talkline. The memory of the JOX era was still around in-house: many of the colleagues who had built the JOX landscape over the years were still in charge.

In 2013, JOX came back. The requirement was concrete: billing and rating should run on JOX again. The decision wasn't made in a crisis, but as a conscious choice based on their own experience. Politically not trivial, technically clear.

Today at freenet DLS

In 2026, JOX controls a substantial part of the mobile operator's operational billing at freenet DLS GmbH. Since its return in 2013, the platform has been in production there.

Metric	As of 2026
Job templates	~4,000
Spread across groups	77
Areas in production	15
JOX locations	89
Active JOX server processes	73
Jobs executed per day	over 23,000
Users (active in 24h / total)	~20 / over 100

This scale spreads across a technically heterogeneous landscape. JOX repository and job targets today cover PostgreSQL, Oracle, MySQL and other DBMS. The operating system is uniformly Linux; operation is split between on-premises and cloud environments.

The 15 production areas structure the operation cleanly: five tenant-specific areas for rating and billing, one cross-brand area for the processes above, others for adjacent applications. On top of that, a second JOX installation serves as a test repository – for development and QA.

At freenet DLS, JOX is used not only as a scheduling platform for ongoing operation but also as a development tool: new processes are implemented and tested directly in JOX. The 100+ users span developers, analysts, operations staff and business-side colleagues.

Established in the early 2000s, the platform has grown several times over. The fundamental architecture has stayed the same: a central repository, distributed server processes, a clean separation between What (job templates in groups) and When and Where (orders with locations in areas). Exactly what was set up in 2001.

What was created in 2001 as a solution for a concrete Talkline problem today carries the backbone of mobile-operator billing. 25 years of practice. One platform.