Service Virtualization Pricing and Licensing Models

« Back to articles

The renewal quote is higher than last year's, even though nothing about the work itself changed. You added six CI agents. A dozen developers now run the simulator on their own laptops. Two more test environments came online. Each addition increased the number the licence counts, and that number is what the invoice bills.

Service virtualization pricing is set by the licence model, not the headline price. Most estates run two usage patterns at once: long-lived 24/7 environments and disposable per-build containers. Only a model priced by peak concurrent instances fits both.

This article covers the licensing models in a table, a rule of thumb for sizing, and a second table of how eight commercial tools charge.

Service virtualization pricing is set by the licence model

Two teams, same quote, different bill

Suppose two teams are quoted the same annual figure for the same tool. The first is six people authoring simulators against a few shared, always-on environments. The second is a platform team whose CI jobs start a fresh simulator and throw it away minutes later, hundreds of times a day.

On a per-seat licence the first team pays for six names; the second pays for seats no ninety-second container can use. On a per-instance or per-call model the fit reverses: the first team overpays, and the second pays only for what it runs. Same quote, a different bill: the model suited one team, not both.

The models, defined

Before you can compare two quotes, define what each one charges for. The models differ on two questions: what unit the licence counts, and what that unit grows with. The table below answers both for each model and maps it to the estate it suits.

Licensing model Charged by Grows with Best-fit use case When AI agents author the simulators
Per-seat / named user each named author number of people a small, stable authoring team where people are the unit, not machines an agent is not a named person, so there is no seat to count it against
Per-server / per-named-instance each long-lived install number of always-on installs a few stable 24/7 environments with no ephemeral spin-up every simulator an agent starts counts as a new install, and it lives minutes
Per-virtual-service each running simulator number of services running at once a fixed, known set of running simulators that rarely changes the bill grows with every simulator an agent authors and runs
Per-call / per-hit (metered) request volume against a cap traffic low-volume or bursty workloads and small free or dev tiers; strains on always-on 24/7 traffic the bill rises at the agent's pace; every test re-run adds calls against the cap
Concurrent floating peak simultaneous instances (with some vendors, concurrent users) from a shared pool peak parallelism a mix of 24/7 environments and ephemeral CI, container and local instances when the pool counts instances, agent-started ones draw a licence and hand it back, so the pool caps the spend
Capacity (PVU) or consumption tokens processor capacity or consumed tokens hardware size or usage organisations already standardised on that vendor's capacity licensing tokens drain at the agent's pace; processor counts grow only if agents spread onto more hardware
Quote-based enterprise negotiated capacity planned central capacity large central estates with predictable, centrally planned demand fixed negotiated capacity; works if the quote anticipated agent workloads
Open-source / self-hosted no licence fee; you pay in build and operate time engineering time, not instances a capable team on the supported protocols willing to own setup and operation no licence for an agent to count against; the cost stays in engineering time

AI agents break the models that count people

Some simulator authoring is no longer done by a person. An AI coding agent takes a ticket, authors a simulator for the dependency it cannot reach, runs the tests, and repeats on the next branch, at machine pace. A seat licence has no one to count, and a metered bill grows at whatever pace the agents work. With a concurrent floating pool that counts instances, the agents' pace does not move the bill: an agent-started instance draws a licence and hands it back, so the pool caps the spend.

A migration has to license two worlds at once

The two worlds, defined

An estate that has run for a few years almost always contains two kinds of workload, and they pull a licence in opposite directions:

  • The 24/7 world is long-lived and shared: a performance environment and a system-test environment that stay up around the clock for any team to use.
  • The disposable world is ephemeral: a simulator starts inside a CI job and is gone when the job ends, or lives on a laptop for an afternoon.

Most teams on containers and continuous delivery run both at once.

Why each traditional model strains on one of them

Take the two traditional models a central-server tool usually ships with; each fits one world and strains on the other:

  • A per-named-user or per-server licence fits the 24/7 world: fixed installs, fixed names, a bill you can predict. In the disposable world every container spin-up counts as a new install, and nobody licenses a name for a simulator that lived ninety seconds.
  • A per-call metered licence suits bursty containers that answer a few thousand calls and stop. An always-on environment runs against the cap around the clock, and a single overnight load test can use up a month's allowance.

A migration carrying both worlds overpays somewhere whichever model it picks.

Concurrent floating sizes to the overlap

A concurrent floating licence is priced on the one number both worlds share: how many instances run at the same moment. The 24/7 environments contribute their steady handful; the disposable containers contribute only their peak overlap, because a finished build hands its licence back to the pool. You pay for the peak simultaneous count, not the install count.

A migration that split the two worlds across two tools

Here is one that happened. In 2020, a global retail bank moved one of its departments off CA LISA (the product now sold as Broadcom Service Virtualization) and, in its executive's words, "cut our tooling costs in half". The case study puts the saving at 51% on tool costs per year.

The department used WireMock for the HTTP simulation and Traffic Parrot for IBM® MQ, in both its performance-testing and system-testing environments. The team said plainly that neither tool matched CA LISA feature for feature, and judged the feature sets were enough at a lower cost. The 51% is one department's figure, not a promise, but it shows what a migration can recover when the licence model is matched to what the estate actually runs.

When you scale, count the instances you run in parallel, not the seats

Metered models grow the bill with usage

The cost of a licensing model is set by how the estate grows, not by the quote you sign. A per-call licence charges more or throttles as the calls multiply. A per-instance or per-virtual-service licence adds an invoice line for every new container. The number that drives the bill is the number that rises when testing works.

Concurrent floating tracks peak parallelism

A concurrent floating licence breaks that link. Double the laptops and CI agents you install on and the bill does not move unless your peak overlap moves with it. Traffic Parrot states it directly: count the instances you run in parallel, not the people who use them.

flowchart LR accTitle: Concurrent floating licensing across an example estate: installs are many, but only the instances running at the same moment hold a licence accDescr: An illustrative example estate drawn as deployment zones around one central licence server, with every multiple drawn UML-style as stacked shapes and no counts shown. In the developer and QA laptop zone a few laptops are running an instance and each holds a licence from the shared pool, while most are installed but not running, greyed out with a dotted line because they hold no licence. In the CI build agent zone, instances hold licences only during builds, and agents between builds are greyed out because they hold no licence. A zone of always-on Docker/Kubernetes environments runs around the clock, so its licences are held 24/7. Installs are many, but licences go only to the instances running at the same moment. pool[("Central licence server (shared pool)")] subgraph laps["Developer and QA laptops"] lapRun@{ shape: procs, label: "A few running an instance" } lapIdle@{ shape: procs, label: "Most installed, not running" } end subgraph agents["CI build agents"] ciRun@{ shape: procs, label: "Instances running during builds" } ciIdle@{ shape: procs, label: "Agents between builds, not running" } end subgraph envs["Always-on Docker/Kubernetes environments"] k8sRun@{ shape: procs, label: "Instances running around the clock" } end pool -->|"a licence while running"| lapRun pool -->|"a licence during builds"| ciRun pool -->|"licences held 24/7"| k8sRun pool -.-|"no licence held"| lapIdle pool -.-|"no licence held"| ciIdle classDef vsvc fill:#e8f2ff,stroke:#2a6fdb,color:#333333; classDef skip fill:#f4f4f4,stroke:#c9c9c9,color:#8a8a8a,stroke-dasharray:4 3; class lapRun,ciRun,k8sRun vsvc; class lapIdle,ciIdle skip; linkStyle 3,4 stroke:#c9c9c9,stroke-width:1px;
An example estate: many installs across laptops, CI agents and always-on environments, yet only the instances running at the same moment hold a licence from the shared pool.

Some floating licences count users, not instances

Two vendors can both say floating and count different units. IBM's Floating licences cap the number of concurrent users, so a licence follows the person using the product. OpenText's concurrent licences, like Traffic Parrot's, are taken by each running instance and returned when it closes; a CI container with no user signed in still draws a licence. Before you compare two floating quotes, ask which unit draws a licence from the pool: a person or an instance.

Sizing a floating pool: peak parallel runs, not agent count

How do you size the pool without over-buying? Traffic Parrot's own worked example shows how to size it. Ten Jenkins build agents where at most three run the tool at once need licences for the three that overlap, not the ten installed, plus the always-on environments alongside them. Peak is a forecast, and a forecast can be wrong:

  • Size for peak, not average. Size too low and a build has to wait for a free licence or does not get one; size too high and you pay for headroom you rarely touch.
  • Measure rather than estimate. A short pilot on your real pipeline finds the peak, and a vendor who models capacity with you before procurement beats one who quotes a round number.

If you are a test lead or an architect, run the pilot and hand procurement the measured peak, not a guess.

Most enterprise service virtualization is priced by quote, and the models barely overlap

Eight commercial tools, four licence families

Line up how the major commercial tools charge and two things stand out: most publish no list price, and the models barely overlap, so a like-for-like price comparison rarely exists. The table below groups eight commercial tools by how they license, from each vendor's own current documentation.

Tool Licensing model What is counted Public list price Notes
Traffic Parrot Concurrent floating Peak simultaneous instances from a shared pool No; budgetary quote, on-premise Priced by the number of floating licences and the protocols and technologies used; one self-contained download runs on laptop, CI, Docker and Kubernetes
Broadcom Service Virtualization (part of DevTest Solutions) We could not find a published licensing model; purchased via Broadcom direct sales and distributors We could not find a published counting unit We could not find a list price on Broadcom's how-to-buy page (accessed 10 July 2026) Central DevTest Server and Registry; self-hosted, with installer and Kubernetes/OpenShift deployment options
Parasoft Virtualize Free desktop edition, paid desktop subscription, and quote-based enterprise or cloud Request volume (hits per day) on the desktop tiers Partly: Professional Desktop from $99/month billed annually; enterprise on quote Free Edition 1,000 hits/day (HTTP/HTTPS only); Professional Desktop 50,000 hits/day; enterprise and cloud unlimited on quote
SmartBear ReadyAPI Virtualization (VirtServer) SmartBear ID-based (SLM) subscription seats on current versions; the older file-based Fixed User and Floating licences are deprecated (no longer issued or supported since 1 October 2025) A per-user seat, bound to a user rather than a machine; each VirtServer instance needs its own licence unless otherwise agreed with SmartBear We could not find a list price in SmartBear's licensing documentation (accessed 10 July 2026) SmartBear's Docker guide documents running current ID-based (SLM) licences in Docker; under the older file-based scheme, Fixed licences were disabled for Docker images and a Floating licence was required (accessed 10 July 2026)
IBM DevOps Test Virtualization (part of IBM DevOps Test Workbench) Authorized User, Floating, Token, and Processor value unit (PVU) licensing Depends on licence type: Authorized users, concurrent users, tokens, or processor capacity We could not find a list price on IBM's product page (accessed 10 July 2026) Four licence types across the Test Workbench package; formerly IBM Rational
WireMock Cloud Subscription SaaS API calls Free tier published (free, 1,000 calls/month); Enterprise "get a tailored quote" Commercial SaaS, distinct from the free WireMock OSS engine
OpenText Service Virtualization Seat and concurrent (floating) licences per OpenText's documentation Machines (seat licence) or concurrent instances from a floating pool; we could not find published prices We could not find a price on OpenText's marketplace listing (accessed 10 July 2026) Formerly Micro Focus; documentation current at version 26.1
Tricentis Tosca Service Virtualization Subscription; delivered as a capability within Tosca We could not find a separately published counting unit We could not find a list price; Tosca pricing is on request The capability is delivered within Tosca rather than listed as a separate product

All vendor licensing sources in the table above accessed 10 July 2026.

All product and company names are trademarks or registered trademarks of their respective owners. Traffic Parrot is not affiliated with, sponsored by, or endorsed by any of the other vendors listed.

Four families cover the eight. Broadcom and Tricentis publish neither a counting unit nor a price that we could find; you buy through a sales conversation. IBM and OpenText each publish a menu of licence types that includes a floating option. SmartBear, WireMock Cloud and Parasoft sell subscriptions with a published counting unit: a user seat, API-call volume, and metered hits per day on Parasoft's desktop tiers. Traffic Parrot licenses one way only, concurrent floating.

Of the eight, only Parasoft and WireMock Cloud put a number on their own pricing pages. The question is not which vendor counts fairly; it is whose counting unit matches the way your estate runs.

Free options, two kinds: self-hosted OSS and free SaaS tiers

Two more options cost nothing to license, and they are not the same kind of free. The first is self-hosted open source: WireMock and MockServer are both Apache 2.0, so there is no licence fee; you pay in engineering time and give up the vendor support contract. WireMock's core is HTTP and HTTPS, with gRPC available through an official extension; MockServer reaches gRPC, WebSockets and TCP too.

The second is a free SaaS tier, where the cap is what to read closely. WireMock Cloud's Free plan is call-capped at 1,000 calls a month; Postman's Free plan is a single seat with unlimited cloud and local mock servers, limited by users, not calls. Both let you try the metered and the seat models before you pay.

Where Traffic Parrot fits, and why

One model, both worlds

Traffic Parrot licenses on the concurrent floating model, installed on-premise. The price is set by the floating licences and the protocols and technologies you need, not a public list. It ships as one self-contained download that runs the same on a laptop, a CI agent and in Docker or Kubernetes, so one licence model and one build cover both worlds. Pricing is per protocol, so a team that needs only HTTP is not paying for IBM® MQ, JMS, gRPC or FIX and SWIFT field matching over files.

What Traffic Parrot does not do

Three limitations sit alongside that fit:

  • Traffic Parrot is on-premise and quote-based, with no public list price and no free SaaS tier. For a number you can read on a page today, or a free plan to prototype on, WireMock Cloud's hosted Free plan or Parasoft's free desktop edition is the closer fit.
  • If your services stop at HTTP and a five-line WireMock OSS stub already does the job, that stub is the right tool.
  • If your estate is purely 24/7 with no ephemeral instances, a traditional per-server licence already fits it.

Where Traffic Parrot pays for itself is the non-HTTP protocols and the mixed estate that runs the 24/7 world and the disposable world at once.

The retail-bank migration from earlier shows the pattern on one department: keep the free HTTP stub where HTTP is enough, and bring in the floating-licensed tool where the protocols or the usage pattern need it.

Frequently Asked Questions

How is service virtualization priced?

Most commercial service virtualization tools are priced by quote rather than a public list, so the licence model matters more than any headline figure. The common models are per-seat, per-server, per-virtual-service, per-call (metered), concurrent floating, capacity units such as PVU, and quote-based enterprise capacity. Open-source tools such as WireMock and MockServer charge no licence fee.

What is a concurrent floating licence?

A concurrent floating licence charges for the number of instances running at the same moment, drawn from a shared pool, rather than for each install or named user. When an instance finishes, its licence returns to the pool. You can install on every laptop, CI agent and container you like and pay only for peak parallel use. Vendors differ on the counted unit: some pools count running instances, others concurrent users, so ask which before you compare quotes.

Do service virtualization tools publish list prices?

Mostly not. Most major commercial tools publish no list price and route buyers through sales or distributors; you compare licence models, then request a quote. The exceptions are partial: Parasoft publishes a desktop price from $99 a month billed annually, and WireMock Cloud publishes a free tier and quotes its Enterprise plan.

How do I license service virtualization for both 24/7 environments and ephemeral containers?

An estate that runs long-lived 24/7 environments and disposable per-build containers needs one model that prices both. Per-server and per-user licences fit the 24/7 installs but bill every container spin-up; per-call licences suit bursty containers, but always-on traffic counts against the cap around the clock. A concurrent floating licence prices the peak simultaneous instances the two worlds share, so one model covers both.

Which licensing model works best when AI agents author the simulators?

A concurrent floating licence, in the majority of cases. An agent is not a named user, so a seat licence has nothing to count, and metered models bill every call and simulator the agent's loop produces. A floating pool that counts instances charges for the peak number running at once, whoever starts them, and caps the spend.

How is Traffic Parrot priced?

Traffic Parrot uses a concurrent floating licence: you pay for the peak number of simulator instances running at once, plus the protocols you license, such as IBM MQ, JMS and gRPC. Pricing is by budgetary quote, not a public list; one self-contained on-premise download covers laptops, CI agents, Docker and Kubernetes.

Where to go next

First, score your usage pattern

Score your own usage pattern before you score any tool. Estimate the peak concurrent instances across your CI fleet, developer laptops and always-on environments, and note which protocols you actually need. Map that pattern to a licensing model using the first table, then shortlist the vendors whose model matches using the second; for most teams the shortlist collapses to two or three. The wider tool decision is in our article on the best service virtualization tools.

Then involve the vendors early

Most enterprise system-simulation licensing is quote-based, so the number that decides your budget does not exist until a vendor produces it. Involve them before your evaluation criteria are final, and have them model the licence cost against your peak-concurrency estimate. Check whether the quote is an annual subscription or a perpetual licence with a yearly maintenance charge on top. Ask what the renewal uplift is after the first term, and whether the agreement carries a true-up or audit clause when usage exceeds what you bought.

If the floating model looks like the fit, request a quote for Traffic Parrot or download the trial and size a pool against your own pipeline. The trial asks for a work email, not a credit card.

Related reading:

Download the Traffic Parrot free 14-day trial