# Apache Messaging: Queues and Streams for Small Business

ActiveMQ, Artemis and Kafka can keep work moving, but they need an operator. Compare all 11 projects, managed prices and three years of ownership costs.

Author: J.A. Watte
Published: October 30, 2026
Source: https://jwatte.com/blog/apache-messaging-event-streaming-small-business/

---

**ActiveMQ, Artemis and Kafka deserve a conversation when your business already has software that needs reliable message delivery. All eleven projects here need an IT provider or developer who can operate them; none is an owner’s replacement for an inbox or a task list.**

A queue can hold a shipping job while the label printer is unavailable. A stream can keep an ordered history of stock changes so several applications can read it. Publish and subscribe means one application sends an event and interested applications receive it. Those are useful building blocks, but somebody must write the applications on either end.

Start with the failure you can describe. Orders might reach the warehouse twice, deliveries might disappear during an internet outage, or three systems might each have a different stock count. A broker can help with the handoff. It cannot decide which system owns the correct answer. For a company using ordinary cloud applications, their existing integration or a managed queue will often be the cheaper first step.

October 30, 2026 update: added setup time and costs when an AI agent does the typing.

## Every project in this group

The table preserves the series catalog’s IT classifications. Dates marked archive date describe distribution files, rather than a confirmed release announcement. We rechecked the three main choices on October 29, 2026: ActiveMQ’s download page says October 27 for 6.4.0, and Kafka’s says October 22 for 4.4.0, superseding earlier catalog date evidence.

| Project | What it does | Who runs it | Latest release and date |
|---|---|---|---|
| [Apache ActiveMQ](http://activemq.apache.org/) | Queues messages so business applications can exchange work reliably. | Your IT department or provider | [6.4.0; 2026-10-27](https://activemq.apache.org/components/classic/download/) |
| [Apache Artemis](https://artemis.apache.org) | Queues and routes messages between business applications. | Your IT department or provider | [2.57.0; 2026-09-09](https://artemis.apache.org/components/artemis/download/) |
| [Apache BifroMQ (Incubating)](https://bifromq.incubator.apache.org/) | Connects many devices and applications through a shared MQTT messaging service. | Your IT department or provider | [4.0.0-incubating; 2026-01-28](https://github.com/apache/bifromq/releases/tag/v4.0.0-incubating) |
| [Apache BookKeeper](http://bookkeeper.apache.org/) | Keeps replicated event logs that other distributed systems depend on. | Your IT department or provider | [4.18.1; 2026-09-16](https://github.com/apache/bookkeeper/releases/tag/release-4.18.1) |
| [Apache EventMesh](https://eventmesh.apache.org) | Routes events between applications and messaging systems. | Your IT department or provider | [1.12.0; 2025-12-29](https://eventmesh.apache.org/download/) |
| [Apache Iggy](https://iggy.apache.org) | Stores and delivers message streams for applications. | Your IT department or provider | [0.9.0; 2026-09-18](https://iggy.apache.org/downloads/) |
| [Apache Kafka](https://kafka.apache.org/) | Carries streams of business events between applications and data systems. | Your IT department or provider | [4.4.0; 2026-10-22](https://kafka.apache.org/community/downloads/) |
| [Apache Pulsar](https://pulsar.apache.org/) | Stores and delivers messages and event streams across applications. | Your IT department or provider | [5.0.1; 2026-10-26](https://pulsar.apache.org/download/) |
| [Apache Qpid](http://qpid.apache.org) | Provides messaging servers and client tools using standard messaging protocols. | Your IT department or provider | [Broker-J 10.1.1; 2026-09-20 (archive date)](https://downloads.apache.org/qpid/broker-j/10.1.1/) |
| [Apache Rocketmq](https://rocketmq.apache.org) | Delivers messages and events between distributed applications. | Your IT department or provider | [5.5.1; 2026-08-20](https://github.com/apache/rocketmq/releases/tag/rocketmq-all-5.5.1) |
| [Apache Vysper](http://mina.apache.org/vysper-project) | Hosts instant messaging services using the XMPP protocol. | Your IT department or provider | [0.7; 2012-12-26 (archive date)](https://downloads.apache.org/mina/vysper/0.7/) |

Apache BifroMQ is incubating. The [ASF explanation](https://incubator.apache.org/guides/branding.html) says that this means it has not yet been fully endorsed by the foundation; incubation alone does not establish whether its code suits your production work.

## ActiveMQ Classic: keep an existing application’s work moving

[ActiveMQ Classic](https://activemq.apache.org/components/classic/documentation/getting-started) is a message broker. A producer submits work and a consumer processes it later. It is a sensible candidate when the company that maintains your Java application already supports it. Choosing it because the name is familiar, before checking that application, reverses the decision.

For example, a nine person HVAC contractor might say that completed jobs sit on tablets until the office software is available again. Its developer could put completed job events into a persistent queue and have the invoicing integration acknowledge them after successful processing. On a normal afternoon, the dispatcher sees pending jobs instead of asking technicians to send them again. Our bounded single-server case below costs $116 to $253/month plus $272 to $817 setup. The offline tablet behavior still requires application code; a broker alone does not supply it.

For example, a twelve person print shop might need order files to wait while a finishing station is offline. Its existing order application could queue a reference to the file, with an order number and a retry count. The production supervisor gets one pending-work view and a separate list for jobs that failed repeatedly. Use the same cost range only if the workload fits the same server; large artwork stays in file storage and needs its own budget. Neither illustration is a reported customer result.

The official getting-started page includes old 5.x disk figures and old operating systems. We would not use its roughly 60 MB distribution figure to size a current production server. The [current download matrix](https://activemq.apache.org/components/classic/download/) lists 6.4.0 and Java compatibility starting at 17. The project offers binary archives and documents Docker. It does not give a universal current RAM minimum on those pages. Our allowance is one Medium server with 8 GiB RAM and 160 GiB disk, pending a test of message sizes, persistence and backlog growth.

You need someone who understands Java updates, TLS, access permissions, persistent messages and recovery. This single broker has an outage window and no cluster failover. A second server alone would not settle failover or storage design. Before using it for paid orders, stop a consumer during a test, restart the broker and prove that unfinished work remains visible. Make processing safe to repeat so one retry cannot send two shipments.

[Amazon MQ pricing](https://aws.amazon.com/amazon-mq/pricing/), read October 29, gives a US East (N. Virginia) example with mq.m5.large ActiveMQ at $0.288 per broker hour and EBS at $0.10 per GB/month. We model one broker, 730 hours/month and its 200 GB storage allocation: 730 × $0.288 + 200 × $0.10 = $230.24/month. This is hourly, without an annual commitment, and excludes transfer and optional network charges. Processor performance and support differ between these service-cost options.

Our self hosted assumption is 4 to 12 setup hours, 1 to 3 administration hours/month and 50 to 100 GB of retained off site copies. Both examples above need separate quotes for application work. A managed service also needs application administration; the table isolates the broker infrastructure decision.

| ActiveMQ single server | Low labor/workload | High labor/workload |
|---|---:|---:|
| Setup labor | $272 | $817 |
| Monthly self hosted | $116 | $253 |
| Self hosted, 36 months | $4,462 | $9,921 |
| Amazon MQ, monthly | $230 | $230 |
| Amazon MQ, 36 months | $8,289 | $8,289 |
| Self hosting savings, 36 months | $3,827 | Negative $1,633 |
| Break even month | 3 | Never |

The low case reaches break even in month 3; the high case never does. That spread is the reason to price the person as carefully as the virtual machine. Skip self hosting if the application vendor requires its own managed service, if nobody can restore the queue, or if an outage stops the whole shop. Begin with the official download and getting-started pages, then ask your application maintainer for its supported broker version and recovery procedure.

## Artemis: a broker for a new integration, with an operator attached

[Apache Artemis](https://artemis.apache.org/components/artemis/documentation/latest/using-server.html) is a separate project and broker. Do not treat an ActiveMQ Classic configuration as an Artemis installation recipe. Its official distribution creates a broker instance with its own configuration and data; the [download page](https://artemis.apache.org/components/artemis/download/) lists 2.57.0, September 9, 2026, and Java 17 or later.

For example, a seven person online retailer might need a packaging job to survive a temporary outage of its label service. A developer could create one queue for accepted packaging requests and another for messages that keep failing. The morning change is practical: staff clear exceptions instead of checking every order in two screens. Our small single-server allowance is $92 to $229/month plus $272 to $817 setup, excluding the connector.

For example, a six person bakery supplying three cafes might want tomorrow’s wholesale orders to reach the kitchen schedule even when its ordering application is being updated. Artemis could hold order events while the kitchen integration reconnects. The baker still needs an obvious warning if the connection has stopped. The same cost allowance applies to a similarly small queue; it does not pay for a new ordering app or an unattended promise that every order is correct.

There is no universal RAM floor in the installation and [performance guidance](https://artemis.apache.org/components/artemis/documentation/latest/perf-tuning.html) we read. The latter explains that memory needs depend on queue sizes and that paging allows work to move to disk. We assume one Small server with 4 GiB RAM for a limited pilot that can tolerate downtime. Test durable storage and a realistic accumulated backlog before approving that size. A Docker image or extracted archive makes a process easier to start, while security and recovery remain IT work.

Ask the developer to explain acknowledgements in business terms: when does the sender consider the request accepted, when does the worker consider it finished, and where can the office see a failed request? Also specify expiry. An old price-change event may become misleading if applied after a newer one. These choices belong in the application contract, regardless of the broker selected.

For a new integration that does not require Artemis protocols, [Amazon SQS](https://aws.amazon.com/sqs/pricing/) is a credible alternative. Its pricing page gives all customers one million requests free each month. Our illustrative low and high workloads are 100,000 to 500,000 total billable request units/month, with messages small enough to avoid extra size units, inside that allowance. Sends, receives, deletes and polling all consume the budget. The queue-service line is therefore $0 in both cases; application hosting, transfer and development are extra. SQS is not a drop-in JMS broker, and changing an existing integration may erase this advantage.

| Artemis single server | Low labor/workload | High labor/workload |
|---|---:|---:|
| Setup labor | $272 | $817 |
| Monthly self hosted | $92 | $229 |
| Self hosted, 36 months | $3,588 | $9,045 |
| SQS request allowance, monthly | $0 | $0 |
| SQS request allowance, 36 months | $0 | $0 |
| Self hosting savings, 36 months | Negative $3,588 | Negative $9,045 |
| Break even month | Never | Never |

Self hosting never breaks even against that request allowance. Choose Artemis because the application needs its capabilities, a reviewed local operating requirement justifies it, or a supported application already uses it. Skip it when a managed queue meets the contract with less operational work. Start with the project’s broker-instance instructions, using a supported Java runtime and a private test network. Do not carry demonstration credentials into production.

## Kafka: useful event history when several systems need it

Kafka keeps event streams that applications can read at their own pace. A consumer can return to retained events and process them again. That differs from treating every event as a one-off task. A business needs a reason for retaining and replaying that history before accepting the extra machinery.

For example, a twenty person delivery company might need dispatch, customer notifications and its daily reports to read the same vehicle-status events. Its provider could use separate consumers, so a broken reporting job does not require the dispatch application to stop. On a normal day, the dispatcher watches delayed updates while the reporting process catches up independently. The six-server planning case below costs $697 to $1,381/month plus $1,634 to $4,085 setup. Vehicle connections and application code are additional work.

For example, a fifteen person auto parts retailer with two locations might want the web store and replenishment report to see stock movements from both counters. A provider could publish recorded stock-change events, keep a bounded history and give each application its own reading position. Staff would investigate one missing event instead of manually comparing complete spreadsheets. Use the same cluster budget only after testing event volume and retention. Kafka does not decide whether an item may be sold twice; the inventory application still controls that rule.

The [official quick start](https://kafka.apache.org/quickstart/) offers downloaded files or an official Docker image and requires Java 17 or later for the local-script route. The [current download page](https://kafka.apache.org/community/downloads/) identifies 4.4.0 as released October 22. A one-process demonstration is useful for learning, but our business example uses separate roles.

The [4.4 KRaft guide](https://kafka.apache.org/44/operations/kraft/) describes three or five controllers and explains why combined broker/controller processes reduce isolation. Three controllers tolerate one controller failure. It gives 5 GB memory and 5 GB disk as a typical controller allowance. Broker resources require separate sizing. Our assumption is three controller servers plus three broker servers, each on the Medium tier, with replication across the three brokers. Broker count and 8 GiB sizing are our bounded workload choices; Apache does not prescribe them for every installation. The [hardware guidance](https://kafka.apache.org/44/operations/hardware-and-os/) makes clear that disk behavior, memory, file descriptors and operating-system configuration matter. A provider must test them and price additional disk capacity if retention needs it.

Our low and high cases retain 100 to 500 GB of off site backups, allow 24 to 60 setup hours and 6 to 16 administration hours/month. Replication is not a substitute for an independent recovery copy or an export plan. An accidental deletion can propagate.

[Confluent Cloud pricing](https://www.confluent.io/pricing/), read October 29, lists Basic’s first eCKU free, then $0.14 per eCKU-hour, data in and out at $0.05/GB and storage at $0.08/GB-month. Our light workload assumes it stays within the first unit: 10 to 100 GB ingress, the same amount of egress and 10 to 100 GB average stored. That makes (10 + 10) × $0.05 + 10 × $0.08 = $1.80/month low, or $18 high. Confluent defines its billed GB as 2^30 bytes; the workload uses that vendor unit, while B2 backup arithmetic uses decimal GB. Paid connectors, extra compute, networking and support are excluded. These usage charges vary with consumption. The service and cluster also have different availability designs.

| Kafka six servers | Low labor/workload | High labor/workload |
|---|---:|---:|
| Setup labor | $1,634 | $4,085 |
| Monthly self hosted | $697 | $1,381 |
| Self hosted, 36 months | $26,732 | $53,796 |
| Confluent Basic usage, monthly | $2 | $18 |
| Confluent Basic usage, 36 months | $65 | $648 |
| Self hosting savings, 36 months | Negative $26,667 | Negative $53,148 |
| Break even month | Never | Never |

The managed option wins both illustrative cases. Ask for a production configuration and quote before relying on Basic’s allowance; the page’s Standard offer has different features and charges. Skip Kafka entirely when a scheduled export or one durable queue solves the actual problem. If replayable history is justified, start with the official quick start using disposable sample events, then have the provider specify retention, access control, consumer lag alerts and a recovery test.

## How we counted the cost

The shared method uses $68.09 per administration hour, an employee compensation estimate from [BLS wages](https://data.bls.gov/oesprofile/?areas=INDUSTRY,STATE,MSA&major_group=150000&measure=01&occupation=151244&year=2025) divided by the [70.0% wages share](https://data.bls.gov/timeseries/CMU2020000000000P). An outside IT provider bills more than this proxy, so get a quote. [DigitalOcean](https://www.digitalocean.com/pricing/droplets) supplies our Small server at $24/month (2 vCPU, 4 GiB RAM, 80 GiB disk), Medium at $48 (4 vCPU, 8 GiB, 160 GiB), and Large at $252 (8 dedicated vCPU, 32 GiB, 100 GiB). These are monthly caps without an annual commitment. [Backblaze B2](https://www.backblaze.com/cloud-storage/pricing) costs $6.95 per decimal TB/month after the first 10 GB free. All were checked October 29, 2026.

Setup = hours × $68.09. Monthly self hosted = server charges + max(0, retained backup GB minus 10) × $0.00695 + administration hours × $68.09. The 36 month total adds setup to 36 monthly bills; savings subtract that total from the alternative. Break even is setup divided by positive monthly savings, rounded up; otherwise it is never. Calculations retain cents internally and display whole dollars.

Our assumptions are 4 to 12 setup hours and 1 to 3 administration hours/month for a bounded single server, or 24 to 60 and 6 to 16 for a small cluster. These exclude application development, migration, training, incidents and on call coverage on both sides. Alternative setup is assumed zero incremental infrastructure labor in these examples. Implementation still takes work. Backup volumes include retained copies. We assume no chargeable backup requests or excess transfer; add them if your usage exceeds the provider allowances. Low and high cases are planning ranges to replace with measured requirements.

## Setup time when Claude or ChatGPT does the typing

We keep the ActiveMQ Classic example: one $48 Medium server and 50 to 100 GB of backups. This remains a provider job even with single-server hours. Our assumed hours include prompting, review, testing and corrections, not unattended agent runtime. Coding studies do not validate these Apache setup savings. [66]

| ActiveMQ Classic: paired planning cases | Low | High |
|---|---:|---:|
| Manual setup hours | 4 | 12 |
| AI setup share; hours | 50%; 2 | 80%; 9.6 |
| Manual administration hours/month | 1 | 3 |
| AI administration share; hours/month | 75%; 0.75 | 100%; 3 |
| Manual total, 36 months | $4,462 | $9,921 |
| Claude Pro subscription total, 36 months | $3,913 | $9,958 |
| ChatGPT Plus subscription total, 36 months | $3,913 | $9,958 |
| API pay per token: Claude Opus 5.5 total, 36 months | $3,727 | $9,872 |
| API pay per token: GPT-6 Astra total, 36 months | $3,750 | $10,043 |
| Opus setup/monthly token charges | $0.764 / $0.382 | $6 / $3 |
| Astra setup/monthly token charges | $1.970 / $0.985 | $15 / $7.50 |

Let S/A be manual setup/monthly hours, q/r the AI shares, H monthly hosting plus backups, and T0/Tm setup/monthly tool charges. We use $68.09/hour and round totals only after calculation.

```text
Manual total = S × $68.09 + 36 × (H + A × $68.09)
AI total = S × q × $68.09 + T0 + 36 × (H + A × r × $68.09 + Tm)
Savings = Manual total minus AI total
```

We budget Claude Pro or ChatGPT Plus at $20 for a separate setup month, then $5/month for 36 operating months: a 25% allocation only when other work uses the rest. Without sharing, add $540. Equal prices do not mean equal allowances; check limits and price extra usage. [54, 55, 56]

Our assumed API budgets are 200,000 to 1,000,000 input and 20,000 to 100,000 output tokens for setup; monthly, 100,000 to 500,000 input and 10,000 to 50,000 output. Low input splits 30% ordinary, 10% short cache writes and 60% reads; high input is uncached. Charges include retries and billed reasoning. The kit shows rates and token arithmetic. [63, 64, 65]

Claude Code or Codex can draft broker configuration and queue recovery checks. Desktop Commander, connected to the server, can install Java, edit broker settings and read persistence logs. Claude in Chrome can inspect the provider panel and ActiveMQ console while we supervise. [57, 59, 60, 61] ChatGPT’s supported cloud browser is separate from your Chrome session, not a promised equivalent extension. [62]

Our base uses free local Desktop Commander or Codex CLI. Optional Remote Pro adds $20 in setup and $5/month with the same allocation, or $200 overall. [58]

Use a new or staging server, keep secrets out of chat and approve commands or use a restricted sandbox. Watch browser actions; use subscriptions for supervised supported tools and API keys for custom automation. Review every change and test a queue restore and duplicate-message handling; we still own backups, updates and the decision.

The conclusion stays: the low case beats Amazon MQ’s $8,289 over 36 months; the high case loses. AI typing does not supply broker failover. The high case saves no monthly labor; recurring fees can erase setup savings.

## Needs your IT department or a provider

The remaining projects are worth recognizing when a vendor proposes them. The monthly hours below are our assumptions: 1 to 3 for a deliberately limited single-server evaluation, or 6 to 16 for a small production cluster. They exclude development and incidents. Where a documented quick start has no numerical hardware floor, we say so; we do not turn an example into a production guarantee.

**BifroMQ (Incubating).** This is MQTT infrastructure for connecting devices and applications. Its [quick-install guide](https://bifromq.incubator.apache.org/docs/get_started/quick_install/) starts one Docker container and verifies publishing and subscriptions. That is the smallest demonstrated setup. Production capacity and a RAM floor are unspecified there. Device credentials, network isolation and reconnect behavior need IT review. Allow 1 to 3 hours/month for evaluation and 6 to 16 if the provider builds a cluster. For example, a two-site car wash could collect equipment status through MQTT. [HiveMQ Cloud Starter](https://www.hivemq.com/pricing/) publishes a managed alternative from $0.34/hour plus $0.80/million messages. At our assumed 730 hours and one million messages, that is $249/month before extras, without an annual term. Protocol compatibility and throughput limits still need checking.

**BookKeeper.** This stores replicated logs for other systems; it is not the queue a receptionist uses. Its [local guide](https://bookkeeper.apache.org/docs/getting-started/run-locally/) runs several bookies for learning, and its [deployment guide](https://bookkeeper.apache.org/docs/deployment/manual/) adds metadata coordination and separate bookie storage. A provider must size that stack and its quorum before claiming a production minimum. Budget our cluster range of 6 to 16 hours/month. A regional delivery firm might meet it underneath Pulsar. Confluent Basic’s published $0.05/GB traffic and $0.08/GB-month storage provide an alternative event-service budget. Changing the underlying storage architecture requires a separate assessment. Compare the parent service rather than buying a log subsystem alone.

**EventMesh.** This routes events between applications and messaging backends. The [current quick start](https://eventmesh.apache.org/docs/quickstart/getting-started/) offers a Docker runtime with an in-memory backend explicitly limited to development and testing; a real deployment adds Kafka or RocketMQ. The smallest trial is one runtime, while production inherits the backend’s requirements. Treat current online documentation separately from the table’s 1.12.0 release and confirm version compatibility. A twelve person manufacturer linking an existing order system to warehouse events might justify it. Allow 6 to 16 hours/month for the combined stack. SQS’s one million free request units/month is a priced alternative for a simpler newly written integration, with application changes required.

**Iggy.** This supplies a streaming server and client libraries. Its [getting-started guide](https://iggy.apache.org/docs/introduction/getting-started/) uses an official 0.9.0 Docker image and requires Linux kernel 5.19 or later. The smallest example is one server; no business workload RAM floor is established there. Container permissions, persistence and client behavior need a developer and operator. A ten person print shop with a custom status feed might evaluate it, allowing 1 to 3 administration hours/month for the pilot. Confluent Basic’s usage rates above are the managed streaming comparison; application and protocol differences prevent a simple server swap.

**Pulsar.** A broker and storage system work together to deliver messages and streams. The [5.0 standalone guide](https://pulsar.apache.org/docs/5.0.x/getting-started-standalone/) runs a learning environment in one process, and requires a 64-bit Java 21 or later runtime for servers. That is not a production availability design or a universal RAM minimum. A provider needs to scope brokers, BookKeeper storage and coordination separately; assume 6 to 16 administration hours/month for a small cluster. A two-location ecommerce business with several independent consumers might justify a review. Confluent Basic is a priced event-stream alternative using the rates above, subject to a client migration assessment because its protocols differ.

**Qpid.** The catalog’s release refers to Broker-J; other Qpid components have separate versions. Its [installation documentation](https://qpid.apache.org/releases/qpid-broker-j-10.0.1/book/Java-Broker-Installation.html) describes a Java broker installed from an archive. Use the matching current manual before selecting the runtime; a one-broker trial does not establish failover capacity. A distributor whose vendor already requires AMQP might use it. Allow 1 to 3 hours/month for a bounded single server, more for a cluster. Amazon MQ’s ActiveMQ price example above provides a published broker-service comparison, but the vendor must approve protocol features, authentication and application behavior before substitution.

**RocketMQ.** Its [local quick start](https://rocketmq.apache.org/docs/quickStart/01quickstart/) starts a NameServer, broker and proxy, then sends and receives test messages. That identifies a small development topology. Production memory still needs sizing. The operator must size storage and plan recovery across those roles. A growing online retailer with software already built for RocketMQ might justify it; budget 6 to 16 administration hours/month for a small production cluster. SQS’s published free request allowance is an alternative for a simpler new queue workload, with a separate rewrite estimate. Do not price only the broker and forget the applications.

**Vysper.** This is an XMPP messaging server, a different job from Kafka. Its [user guide](https://mina.apache.org/vysper-project/user_guide.html) is sparse and the catalog’s 0.7 archive dates to 2012. A current supported Java and hardware baseline was not verified, so we would not start a new business deployment from it. A nonprofit with a legacy XMPP integration should have its provider inventory the dependency and price migration. Allow at least our 1 to 3 hours/month planning range for a legacy single server, without pretending that buys security support. [Slack Pro](https://slack.com/pricing) lists $8.75 per person/month billed monthly, or $7.25 with annual billing, before temporary promotions. It can replace the staff-chat job, with any XMPP integrations separately migrated.

## Your developer’s toolbox

There are no developer-tier entries in this group’s catalog. That does not mean development is optional. The broker stores and delivers an event; your software gives it a meaning. Specify a unique business identifier, a supported event format and what happens when the same event arrives twice. Put an owner on failed messages, and keep a visible reconciliation report. A green server indicator does not prove that yesterday’s orders were invoiced.

## Risks and support

Keep broker administration screens private. Give each application only the destinations it needs, encrypt connections and remove demonstration accounts. Review the [ActiveMQ advisories](https://activemq.apache.org/security-advisories), [Artemis advisories](https://artemis.apache.org/security-advisories) and [Kafka security notices](https://kafka.apache.org/community/cve-list/) and [ASF security guidance](https://www.apache.org/security/). Follow the selected project’s release and security channels as part of the monthly work. Never publish customer payloads in a public bug report.

A recent release is useful evidence, but read the release notes and mailing lists too. A long silence, unsupported runtime or movement to the [Apache Attic](https://attic.apache.org/) warrants a replacement plan. Vysper’s old archive illustrates why catalog membership is not an endorsement for a new installation. Ask an existing vendor who supplies patches, what versions it supports and how an urgent fix is delivered.

Amazon MQ, Confluent and HiveMQ publish the commercial services linked above. Their service contracts define what they support; do not assume they support your separately hosted Apache installation. For self hosting, get a written scope covering updates, recovery and the applications, with response times priced separately.

The [Apache License 2.0](https://www.apache.org/licenses/LICENSE-2.0) permits commercial use and modification under its conditions, including required notices when redistributing. It provides no warranty and does not grant rights to use Apache trademarks as your own. Maintenance requires a separate agreement.

## Download the kit

The [messaging and event streaming kit](/downloads/apache-messaging-event-streaming-small-business-kit.md) includes every project, provider questions, a first-week evaluation and the same cost worksheets with blanks for your workload.

## Related reading

- [The Apache small business series](/blog/apache-open-source-projects-small-business/).
- [Workflow automation and integration](/blog/apache-workflow-automation-small-business/).
- [Operational databases](/blog/apache-databases-small-business/).
- [Device and machine data](/blog/apache-iot-edge-devices-small-business/).
- [n8n for small business automation](/blog/n8n-small-business-automation-guide/).

## Fact-check notes and sources

Official pages and the series catalog evidence were read October 29, 2026. Prices are US dollars before tax; billing terms appear with each comparison. We have not installed these projects for this series.

1. [Apache ActiveMQ](http://activemq.apache.org/). Read October 29, 2026, 21:12:49 UTC (catalog observation).
2. [6.4.0; 2026-10-27](https://activemq.apache.org/components/classic/download/). Read October 29, 2026, 21:39:20 UTC.
3. [Apache Artemis](https://artemis.apache.org). Read October 29, 2026, 21:12:52 UTC (catalog observation).
4. [2.57.0; 2026-09-09](https://artemis.apache.org/components/artemis/download/). Read October 29, 2026, 21:37:53 UTC.
5. [Apache BifroMQ (Incubating)](https://bifromq.incubator.apache.org/). Read October 29, 2026, 21:38:47 UTC.
6. [4.0.0-incubating; 2026-01-28](https://github.com/apache/bifromq/releases/tag/v4.0.0-incubating). Read October 29, 2026, 21:19:11 UTC (catalog observation).
7. [Apache BookKeeper](http://bookkeeper.apache.org/). Read October 29, 2026, 21:38:46 UTC.
8. [4.18.1; 2026-09-16](https://github.com/apache/bookkeeper/releases/tag/release-4.18.1). Read October 29, 2026, 21:22:49 UTC (catalog observation).
9. [Apache EventMesh](https://eventmesh.apache.org). Read October 29, 2026, 21:38:47 UTC.
10. [1.12.0; 2025-12-29](https://eventmesh.apache.org/download/). Read October 29, 2026, 21:13:25 UTC (catalog observation).
11. [Apache Iggy](https://iggy.apache.org). Read October 29, 2026, 21:13:38 UTC (catalog observation).
12. [0.9.0; 2026-09-18](https://iggy.apache.org/downloads/). Read October 29, 2026, 21:14:12 UTC (catalog observation).
13. [Apache Kafka](https://kafka.apache.org/). Read October 29, 2026, 21:13:39 UTC (catalog observation).
14. [4.4.0; 2026-10-22](https://kafka.apache.org/community/downloads/). Read October 29, 2026, 21:39:18 UTC.
15. [Apache Pulsar](https://pulsar.apache.org/). Read October 29, 2026, 21:13:44 UTC (catalog observation).
16. [5.0.1; 2026-10-26](https://pulsar.apache.org/download/). Read October 29, 2026, 21:15:28 UTC (catalog observation).
17. [Apache Qpid](http://qpid.apache.org). Read October 29, 2026, 21:13:44 UTC (catalog observation).
18. [Broker-J 10.1.1; 2026-09-20 (archive date)](https://downloads.apache.org/qpid/broker-j/10.1.1/). Read October 29, 2026, 21:25:11 UTC (catalog observation).
19. [Apache Rocketmq](https://rocketmq.apache.org). Read October 29, 2026, 21:13:45 UTC (catalog observation).
20. [5.5.1; 2026-08-20](https://github.com/apache/rocketmq/releases/tag/rocketmq-all-5.5.1). Read October 29, 2026, 21:16:03 UTC (catalog observation).
21. [Apache Vysper](http://mina.apache.org/vysper-project). Read October 29, 2026, 21:38:48 UTC.
22. [0.7; 2012-12-26 (archive date)](https://downloads.apache.org/mina/vysper/0.7/). Read October 29, 2026, 21:25:11 UTC (catalog observation).
23. [ASF explanation](https://incubator.apache.org/guides/branding.html). Read October 29, 2026, 21:37:59 UTC.
24. [ActiveMQ Classic](https://activemq.apache.org/components/classic/documentation/getting-started). Read October 29, 2026, 21:37:52 UTC.
25. [Amazon MQ pricing](https://aws.amazon.com/amazon-mq/pricing/). Read October 29, 2026, 21:38:33 UTC.
26. [Apache Artemis](https://artemis.apache.org/components/artemis/documentation/latest/using-server.html). Read October 29, 2026, 21:37:52 UTC.
27. [performance guidance](https://artemis.apache.org/components/artemis/documentation/latest/perf-tuning.html). Read October 29, 2026, 21:38:46 UTC.
28. [Amazon SQS](https://aws.amazon.com/sqs/pricing/). Read October 29, 2026, 21:38:53 UTC.
29. [official quick start](https://kafka.apache.org/quickstart/). Read October 29, 2026, 21:37:52 UTC.
30. [4.4 KRaft guide](https://kafka.apache.org/44/operations/kraft/). Read October 29, 2026, 21:38:47 UTC.
31. [hardware guidance](https://kafka.apache.org/44/operations/hardware-and-os/). Read October 29, 2026, 21:40:16 UTC.
32. [Confluent Cloud pricing](https://www.confluent.io/pricing/). Read October 29, 2026, 21:38:42 UTC.
33. [BLS wages](https://data.bls.gov/oesprofile/?areas=INDUSTRY,STATE,MSA&major_group=150000&measure=01&occupation=151244&year=2025). Read October 29, 2026, 21:12:08 UTC.
34. [70.0% wages share](https://data.bls.gov/timeseries/CMU2020000000000P). Read October 29, 2026, 21:17:59 UTC.
35. [DigitalOcean](https://www.digitalocean.com/pricing/droplets). Read October 29, 2026, 21:10:42 UTC.
36. [Backblaze B2](https://www.backblaze.com/cloud-storage/pricing). Read October 29, 2026, 21:10:42 UTC.
37. [quick-install guide](https://bifromq.incubator.apache.org/docs/get_started/quick_install/). Read October 29, 2026, 21:40:16 UTC.
38. [HiveMQ Cloud Starter](https://www.hivemq.com/pricing/). Read October 29, 2026, 21:37:58 UTC.
39. [local guide](https://bookkeeper.apache.org/docs/getting-started/run-locally/). Read October 29, 2026, 21:40:17 UTC.
40. [deployment guide](https://bookkeeper.apache.org/docs/deployment/manual/). Read October 29, 2026, 21:39:19 UTC.
41. [current quick start](https://eventmesh.apache.org/docs/quickstart/getting-started/). Read October 29, 2026, 21:39:18 UTC.
42. [getting-started guide](https://iggy.apache.org/docs/introduction/getting-started/). Read October 29, 2026, 21:37:57 UTC.
43. [5.0 standalone guide](https://pulsar.apache.org/docs/5.0.x/getting-started-standalone/). Read October 29, 2026, 21:39:20 UTC.
44. [installation documentation](https://qpid.apache.org/releases/qpid-broker-j-10.0.1/book/Java-Broker-Installation.html). Read October 29, 2026, 21:40:30 UTC.
45. [local quick start](https://rocketmq.apache.org/docs/quickStart/01quickstart/). Read October 29, 2026, 21:37:57 UTC.
46. [user guide](https://mina.apache.org/vysper-project/user_guide.html). Read October 29, 2026, 21:40:30 UTC.
47. [Slack Pro](https://slack.com/pricing). Read October 29, 2026, 21:37:59 UTC.
48. [ActiveMQ advisories](https://activemq.apache.org/security-advisories). Read October 29, 2026, 21:37:54 UTC.
49. [Artemis advisories](https://artemis.apache.org/security-advisories). Read October 29, 2026, 21:39:20 UTC.
50. [Kafka security notices](https://kafka.apache.org/community/cve-list/). Read October 29, 2026, 21:40:17 UTC.
51. [ASF security guidance](https://www.apache.org/security/). Read October 29, 2026, 21:39:57 UTC.
52. [Apache Attic](https://attic.apache.org/). Read October 29, 2026, 22:09:30 UTC.
53. [Apache License 2.0](https://www.apache.org/licenses/LICENSE-2.0). Read October 29, 2026, 21:39:56 UTC.

54. [Claude plans](https://claude.com/pricing). Read October 29, 2026, 23:31:08 UTC; reused for this update.
55. [ChatGPT plans](https://chatgpt.com/pricing/). Read October 29, 2026, about 23:27 UTC in a regular browser.
56. [Codex plan access](https://help.openai.com/en/articles/11369540-using-codex-with-your-chatgpt-plan). Read October 29, 2026, 23:31:16 UTC; reused for this update.
57. [Codex CLI](https://developers.openai.com/codex/cli). Read October 29, 2026, 23:32:01 UTC; reused for this update.
58. [Desktop Commander pricing](https://desktopcommander.app/pricing/). Read October 29, 2026, 23:31:08 UTC; reused for this update.
59. [Desktop Commander setup and permissions](https://raw.githubusercontent.com/wonderwhy-er/DesktopCommanderMCP/main/README.md). Read October 29, 2026, 23:31:41 UTC; reused for this update.
60. [Claude in Chrome](https://support.claude.com/en/articles/12012173-get-started-with-claude-in-chrome). Read October 29, 2026, 23:31:11 UTC; reused for this update.
61. [Chrome permissions](https://support.claude.com/en/articles/12902446-claude-in-chrome-permissions-guide). Read October 29, 2026, 23:31:09 UTC; reused for this update.
62. [ChatGPT cloud browser](https://help.openai.com/en/articles/20001280-using-cloud-browser-in-chatgpt). Read October 29, 2026, 23:32:01 UTC; reused for this update.
63. [Claude API pricing](https://platform.claude.com/docs/en/about-claude/pricing). Read October 29, 2026, 23:31:14 UTC; reused for this update.
64. [OpenAI API pricing](https://developers.openai.com/api/docs/pricing). Read October 29, 2026, 23:31:13 UTC; reused for this update.
65. [OpenAI prompt caching](https://developers.openai.com/api/docs/guides/prompt-caching). Read October 29, 2026, 23:32:00 UTC; reused for this update.
66. [METR coding study limitations](https://metr.org/blog/2026-02-24-uplift-update/). Read October 29, 2026, 23:31:09 UTC; reused for this update.

*Apache, the Apache feather logo and the Apache project names are trademarks of The Apache Software Foundation. This site is not affiliated with or endorsed by the ASF. Other product names are used for identification only.*


---

Canonical HTML: https://jwatte.com/blog/apache-messaging-event-streaming-small-business/
RSS: https://jwatte.com/feed.xml
JSON Feed: https://jwatte.com/feed.json
Hero image: https://jwatte.com/images/apache-messaging-event-streaming-small-business.webp