SaaS is a simple way to get started with API Connect. There’s nothing to deploy, no infrastructure to build out and no upgrades to plan - you sign up and start building and publishing APIs. For a lot of teams that’s exactly what they want.
It’s not right for everyone though. I’ve seen teams pick SaaS because it was the quickest way to start, only to hit a requirement later that it was never designed to meet. I’ve also seen teams move from one option to another expecting a like-for-like swap, and only discover afterwards what they’d left behind.
The comparison page shows what each option includes. This post is the companion to it: what those differences mean in practice, what you give up when you move, and how to work out which option fits you rather than which one looks easiest on day one. Use the two together, and check the page for current details before you make a final decision.
The options
There are three main ways to run API Connect, and each is the right answer for someone.
IBM Integration SaaS - the management plane is shared with other customers, and there is a faster rate of change with features coming ahead of the software releases. You have a dedicated gateway cluster, and remote gateways are available if you want traffic to stay close to your backend systems. IBM runs and upgrades the service on a schedule we plan. This is hosted on AWS or Azure.
Reserved Instance - a single-tenant environment that we still run for you, but with some say in upgrade scheduling and a lot more flexibility in what you can configure. It’s hosted on IBM Cloud.
Software - you deploy and run API Connect yourself, wherever you need it: on-premises, in your own cloud accounts, or in environments that can’t reach the internet at all. You decide when to upgrade, how it’s configured, how it fits with your existing operations tooling and where every piece of data lives. You also take on running it - capacity, upgrades, backups, monitoring and on-call.
Software isn’t a fallback for when SaaS doesn’t fit. For many organisations it’s the best choice from the start, particularly if you already have a platform team running other workloads alongside your APIs.
There is also a pay-as-you-go SaaS offering, which is the quickest to start with and has a narrower set of capabilities than the three above e.g. Nano Gateway only, no remote gateways.
Moving between options
Moving from Software to Integration SaaS
This is the move we get asked about most, and often the reason is simply “we don’t want to run it any more”. That’s a perfectly good reason. The patching, upgrades, monitoring, backups, capacity planning and on-call all go away, and for a lot of teams that’s worth a great deal.
What you hand over in exchange is control, and it’s worth being specific about where.
The gateway. There’s no DataPower console in Integration SaaS, and there isn’t yet a way to get at the gateway logs, whereas Reserved Instance provides them through IBM Cloud Logging. If your operations teams rely on either for troubleshooting, you’ll need to work out what they’ll use in the meantime. The v5c gateway isn’t available either, so if that’s what you’re running today you’ll have to move to a different gateway before you move to SaaS, which is a piece of work in its own right.
Limits. The management plane is shared with other customers, so there are limits that you won’t have had when you owned the whole thing - on the number of catalogs and user registries, for example, and on how long analytics data is kept. If you have lots of catalogs, or you keep analytics for audits or capacity planning, validate this up front.
Timing of change. Updates are applied on IBM’s schedule, not yours. You get fixes and features quickly, but you can’t hold a release back for your financial year end or peak trading period.
Availability and recovery. You get the service’s SLA and recovery objectives, not ones you’ve engineered yourself. If your own targets are tighter, compare them honestly.
Where it runs. Integration SaaS is hosted on AWS or Azure in the regions IBM offers. If your data residency rules, certifications or internal policies require somewhere else, that may settle the question on its own.
Your backends. If your backends are on-premises or in a private network, the gateway in SaaS has to reach them. That might mean opening up connectivity you’ve never needed before, or running a remote gateway close to them - which works well, but it does mean you’re still running something yourself.
IP allow lists. If your backends or firewalls only accept traffic from allow-listed IP addresses, these will need to be updated based on the published source IPs available.
Access control. Our hosted offerings use IBMid for access to API Manager - with Reserved Instance integrating with IBM Cloud IAM and Integration SaaS having platform-wide role configuration. This will be different from what you use in Software and you may need to set up IBMid Enterprise Federation to integrate with your user directory.
Private connectivity. Depending on where you run the gateways you will have different network paths to backend services. The private connectivity through Private Link connects the data-plane to your cloud account - reaching to on-premise networks is up to the hyperscaler’s offerings.
CI/CD and automation. The same API Connect Provider API and CLI are available in the hosted offerings, but you’ll need to consider authentication, hostname and possibly routing changes.
Platform management. The managed offerings are run with the model that we handle the infrastructure and platform configuration so you can focus on your APIs. This means that Cloud Manager is not available - but we provide a separate interface to manage remote gateways and analytics offload configuration.
None of that means you shouldn’t do it. If your usage fits comfortably inside these boundaries and running the platform is a cost you’d rather not carry, Integration SaaS can be a really good move. Just make sure you find the boundaries before you cross them rather than after. Make a list of what you use today - gateway features, integrations, retention and so on - and check it against the comparison page. A trial with a representative set of your APIs is better still, and keeping your existing deployment running until you’re confident gives you somewhere to go back to.
Moving from Reserved Instance to Integration SaaS
Most of the points above apply here too: you get a simpler service and less to think about, but Reserved Instance has more flexibility, and that is what you’re giving up. In particular, look at how much you can customise, gateway log access, and the choice over when upgrades happen.
Two things are specific to this move. The offerings are hosted on different platforms (IBM Cloud and AWS or Azure) and cover different regions, so check the region you need is available. And if you use other IBM Cloud services alongside Reserved Instance, there aren’t direct equivalents on the other platforms.
Because these details change over time, treat the comparison page as the source of truth for the current position.
Moving from a managed option to Software
You gain control over upgrade timing, configuration, where it runs and how it fits with everything else you operate. You also take on everything we were doing for you - keeping it patched and upgraded, monitoring it, responding when it breaks, backups and disaster recovery, and the people and skills to do all of that. Make sure that work is planned and staffed before you move, not discovered afterwards.
The move itself
Whichever direction you go, the move itself is a project. APIs, products, subscriptions and credentials all need to come across, and if your API endpoints change then every consumer is affected too. At the time of writing the recommendation for tooling is to set up your instances and catalogs, point your CI/CD at the new environment and then make use of ConfigSync to bring consumers, subscriptions, applications and credentials to the new environment. This is a job that can be run periodically so will likely need multiple runs to align with cut over. If you don’t have CI/CD in place already you can use the API Connect toolkit CLI to do this but it is likely more effort. Migration of analytics data is not possible so ensure you have a plan for any data you need to retain. Plan for testing that you can reuse, not a one-off comparison, and plan how you’d go back if you needed to.
Key questions
There’s no single right answer, but working through these questions before you commit usually makes the choice clearer. The detail behind each one is in the sections above.
- What certifications and compliance standards do you need to meet? This can rule options in or out before anything else is considered. Check which offerings hold the certifications your industry or customers require, and whether the regions you need are covered. With software you can deploy into an environment you’ve already accredited, which may be simpler than getting a new service approved.
- Where do your backends live? Every extra network hop between the gateway and your backends adds latency and another thing to secure. Cloud backends suit a gateway in a nearby region; on-premises or private ones suit a remote gateway close to them, with the management plane run for you. Some private connectivity options depend on other cloud services that are billed separately.
- How available does it need to be, and how will you recover? Compare the SLA and recovery objectives of each option honestly with what your business needs. Think about how separate your development, test and production environments need to be too - for some, separate provider organisations or catalogs are plenty; others need dedicated infrastructure so a noisy neighbour can’t affect production.
- Who decides when things change? If you have times of year when nothing is allowed to change, remember that we update the managed offerings on our schedule. Reserved Instance has some flexibility, but there are times we will need to mandate updates for security fixes.
You don’t have to pick just one
This doesn’t have to be all or nothing. Remote gateways are available in all three options, so where your APIs are managed and where they actually run can be separate decisions. You could have the management plane run for you and put the gateways in your own data centre or cloud account, next to your backends. You could use different options for different needs, such as development and test on one and production on another. And as covered above, you can change your mind later, though it’s a project.
If you want to keep that flexibility, think about which gateway features you’re relying on. The DataPower Nano Gateway is available in every option, so APIs built around it are the easiest to move.
Making the choice
If the answers point to minimal constraints and you’d rather not run the platform, Integration SaaS is a really good option - you get to focus on your APIs instead of what’s running them. If you want a dedicated environment, more flexibility in the gateway and more control over change, but still want it run for you, Reserved Instance is a good fit. If you need to own where it runs, when it changes and how it fits with the rest of your estate, software is the right choice, and it’s worth considering seriously from the start rather than as a last resort.
Whichever you choose, revisit these questions from time to time. And if you’re thinking of moving, weigh up what you’ll lose as carefully as what you’ll gain.