What is a realistic SLA for an Ignition MES or SCADA platform?
A realistic SLA for an Ignition MES or SCADA platform is written in time, not only in percent. It states how fast someone responds and how fast the plant is running again, per severity, 24/7.

A realistic SLA for an Ignition MES or SCADA platform is written in time, not only in percent. It states how fast someone responds and how fast the plant is running again, per severity, 24/7. Percent uptime hides when the downtime happens: 99.9% still allows almost 9 hours a year, which can land in the middle of a shift.
Most service level agreements you will see for industrial software were written for cloud services. They promise a monthly uptime percentage and pay a service credit when they miss it. That model works for a web app. It works badly for a plant where one Ignition gateway runs the screens, the historian and the MES logic for three shifts a day.
This guide is for operations directors, OT and IT managers and procurement leads who are writing or reviewing an Ignition SLA. It explains why a percentage misleads on a plant floor, what an SLA in units of time looks like, which clauses to put in the contract, which architecture choices make tight targets possible, and what to ask a provider before you sign.
Key takeaways
- An Ignition SLA covers the run side of the platform: gateways, databases, servers, network paths and clients, plus the people who respond when they fail.
- 99.9% uptime equals 8.76 hours of allowed downtime a year, or 43.2 minutes in a 30-day month. The percentage says nothing about when that downtime happens.
- A time-based SLA states a response time and a restore time for each severity level, 24/7/365. Those two numbers are what the plant actually feels.
- Response time and restore time are different promises. Many contracts only commit to the first one.
- Tight restore targets need architecture behind them: redundancy, a recovery node, safe staging for patches and no dependency on an outside cloud.
- Ask for planned maintenance rules, exclusions, escalation and exit terms in writing before you compare prices.
How to write a realistic Ignition SLA in 6 steps
- List what is in scope: every gateway, database, server, network segment and client type the plant depends on.
- Define severity levels in plant terms, for example "production stopped" versus "a report is late".
- For each severity, set a response time and a restore time, and state the hours they apply (24/7/365 or business hours).
- Agree how planned maintenance and patching are scheduled around shifts, and whether they count as downtime.
- Write down the exclusions, the escalation path and the reports you will receive.
- Check that the architecture can actually deliver the restore times you agreed.
What is an SLA for an Ignition platform?
An SLA for an Ignition platform is a contract that states which parts of your Ignition system a provider keeps running, how fast they react when something breaks, how fast they restore it, and what happens when they miss. It covers gateways, databases, servers, network paths and clients, and is signed by the plant owner and the provider who operates the stack.
Ignition by Inductive Automation is a platform, not a single application. In one plant it can be the SCADA layer, the MES layer, the historian and the reporting layer at once. That is why "the Ignition SLA" is really a bundle of promises about several moving parts:
- Gateways: the Ignition gateway service, its modules, its licence activation and its connections to PLCs and devices.
- Databases: the SQL database or historian that Ignition writes to, including storage, retention and backups.
- Servers and operating system: the physical or virtual machines, the hardened OS, disks, power supplies and certificates.
- Network paths: the links from the gateway to the control network, to clients and to IT systems such as ERP.
- Clients: Perspective sessions, Vision clients and the panel PCs or thin clients on the line.
Three different SLAs often get confused here. Inductive Automation sells support plans for the Ignition software itself. A system integrator may sign a project warranty. A managed service provider signs an operational SLA for your specific servers. Only the last one promises to get your plant running again at 3 a.m. on a Sunday. If you are not sure which kind you are reading, the clause that tells you is the one that defines "restore".
It also helps to know what the software vendor itself commits to. Inductive Automation's support plan page lists initial response times of 16 hours for BasicCare, 4 hours for TotalCare and 2 hours for PriorityCare, valid during normal support hours, with 24x7x365 emergency support on TotalCare and PriorityCare. Those are response targets for software questions. They are not restore targets for your plant, and they were never meant to be. Your operational SLA has to fill that gap.
Why does a % uptime SLA mislead on a plant floor?
A percentage uptime SLA misleads because it measures an average, while a plant is hurt by single events. 99.9% allows 8.76 hours of downtime a year, and the contract does not care if that is one full night shift or forty short stops. Monthly windows, maintenance exclusions and service credits weaken it further.
Here is what common uptime percentages allow, on a 365-day year (8,760 hours) and a 30-day month (720 hours). The formula is simple: allowed downtime = period length x (100% minus the SLA percentage).
Allowed downtime by uptime SLA
99% uptime
Per year: 87.6 h
Per 30-day month: 7 h 12 min
Per week: 1 h 40.8 min
99.5% uptime
Per year: 43.8 h
Per 30-day month: 3 h 36 min
Per week: 50.4 min
99.9% uptime
Per year: 8.76 h
Per 30-day month: 43.2 min
Per week: 10.1 min
99.95% uptime
Per year: 4.38 h
Per 30-day month: 21.6 min
Per week: 5.0 min
99.99% uptime
Per year: 52.6 min
Per 30-day month: 4.3 min
Per week: 1.0 min
- 99% uptime: 87.6 hours a year, or 7 hours 12 minutes in a 30-day month.
- 99.5% uptime: 43.8 hours a year, or 3 hours 36 minutes a month.
- 99.9% uptime: 8.76 hours a year (8 hours 45.6 minutes), or 43.2 minutes a month.
- 99.95% uptime: 4.38 hours a year (4 hours 22.8 minutes), or 21.6 minutes a month.
- 99.99% uptime: 52.6 minutes a year, or 4.3 minutes a month.
Four things make these numbers look better on paper than they feel in a plant.
1. The percentage ignores timing. An hour of downtime during a planned weekend stop costs almost nothing. The same hour at the start of a shift, during a batch or during a changeover can cost a lot more. A percentage treats both hours the same.
2. The percentage ignores the shape of the outage. 8.76 hours as one event is a lost shift. 8.76 hours as dozens of short drops is a different problem: operators lose trust in the screens, data gaps appear in the historian and people start writing on paper again. Both pass a 99.9% annual target.
3. The measurement window hides things. Most cloud SLAs measure monthly. AWS, for example, measures EC2 availability per monthly billing cycle in its compute SLA. Monthly windows reset the budget every month, and many contracts exclude planned maintenance, customer network issues and third-party failures from the count. Read what counts as "unavailable" before you read the percentage.
4. The remedy is money, not production. When a cloud SLA is missed, the usual remedy is a service credit. In the AWS compute SLA, an availability between 99.0% and just under 99.99% at region level earns a 10% credit on that month's bill, credits go against future payments rather than refunds, and the SLA is described as the sole and exclusive remedy for unavailability. A 10% credit on a hosting bill does not pay for a lost shift.
Dominique Wille, who leads Plantformance, puts it simply: "IT service suppliers give SLAs in percent, mostly for cloud. We can give an SLA in units of time, which is rare."
A worked example: 99.9% uptime vs an SLA in hours
Numbers make the difference clearer than any argument. Take one plant running Ignition for SCADA and MES on three shifts, and look at three things that can happen to it in a year. We compare each against two contracts.
- Contract A: 99.9% uptime. Version A1 measures per year, version A2 per 30-day month. Remedy: a service credit.
- Contract B: an SLA in units of time. Severity 1 ("production stopped or blind") gets a time to first response (TTFR) within R and a restore within T, 24/7/365. For the math below we use R = 15 minutes and T = 4 hours as example values only. They are not Plantformance figures and not a recommendation.
Same outages, two contracts
Scenario 1
One long outage on a night shift
99.9% contract: met
Time-based contract: breached by 4 h 45 min
Scenario 2
Three short stops in one month
99.9% contract: met
Time-based contract: met, repeat-incident rule triggers
Scenario 3
A six-hour outage that trips the percentage
99.9% contract: missed (10% credit)
Time-based contract: breached by 2 h
Scenario 1: one long outage on a night shift
A disk controller fails at 1 a.m. and the gateway is down for 8 hours 45 minutes (525 minutes).
- Annual 99.9% (A1): a year has 525,600 minutes, and 99.9% allows 525.6 of them. 525 minutes of downtime gives (525,600 minus 525) / 525,600 = 99.9001% uptime. The SLA is met. Nobody owes the plant anything, even though a full shift ran blind.
- Time-based (B): the restore target was 4 hours. The outage ran 8 hours 45 minutes. The SLA is breached by 4 hours 45 minutes, and the breach is visible the moment the clock passes 4 hours, not at the end of the year.
Scenario 2: three short stops in one month
The gateway drops three times in a 30-day month, 14 minutes each time, every time near a shift start.
- Monthly 99.9% (A2): the month has 43,200 minutes, and 99.9% allows 43.2. Three stops of 14 minutes are 42 minutes, so uptime is (43,200 minus 42) / 43,200 = 99.903%. The SLA is met, with 1.2 minutes to spare.
- Time-based (B): each stop was restored in 14 minutes, well inside T. The restore target is met too. This is where a good time-based SLA adds one more clause: a repeat-incident rule, for example a mandatory root cause report after the second Severity 1 event on the same component within 30 days. The percentage contract has no way to say "this keeps happening".
Scenario 3: a six-hour outage that trips the percentage
A failed update leaves the gateway unusable for 6 hours in a 30-day month.
- Monthly 99.9% (A2): (720 minus 6) / 720 = 99.17% uptime. The SLA is missed. Under a credit table like the AWS one above, 99.17% falls in the 99.0% to 99.99% band, so the remedy is 10% of that month's fee. The plant lost 6 hours of production; the provider loses a tenth of one invoice.
- Time-based (B): the restore target was 4 hours, so the SLA is breached by 2 hours, and the contract can say what follows: escalation to named people, a root cause report within an agreed time and service changes, not only a credit.
What the example shows: the percentage contract was met in two of three scenarios, including the one where a full shift ran blind. The time-based contract caught both long outages and gave the plant a lever on the repeated short ones. In every scenario, what the plant felt was the restore time, not the percentage.
Need an SLA written in hours, not percent? Talk to the Plantformance team about response and restore times for your Ignition gateways. Book a demo
What does an SLA in units of time look like?
An SLA in units of time lists severity levels and gives each one a response time and a restore time, valid 24/7/365 or within stated hours. It replaces "99.9% per month" with "if production stops, an engineer is working on it within R and the plant is running within T", plus rules for planned maintenance around shifts.
A simple structure looks like this. The letters are placeholders; the real values depend on your plant, your architecture and your provider.
- Severity 1, production stopped or blind: the gateway, database or network path is down, or operators cannot see or control the process. Response within R1, restore within T1, 24/7/365.
- Severity 2, production degraded: the line runs, but a key function is broken, for example historian writes failing, one area's screens down, alarms delayed. Response within R2, restore within T2, 24/7/365.
- Severity 3, limited impact: a report, a dashboard or a non-critical integration fails, with a workaround in place. Response within R3, restore within T3, business hours.
- Severity 4, requests and questions: changes, new users, how-to questions. Response within R4, planned with the plant.
Three details turn that table into something you can hold a provider to.
Restore means the plant runs again, not that the root cause is fixed. A good contract separates restore (production is back, possibly on a workaround or a backup node) from resolve (the root cause is found and fixed for good). Restore is the number operations cares about. Resolve belongs in the root cause report.
The clock is defined. When does it start: when an alarm fires, when the plant calls or when a ticket is logged? When does it stop: when the provider says so or when the plant confirms? Write it down.
Planned work has its own rules. Patching and upgrades happen in maintenance windows agreed with the plant, around shifts and production peaks, announced in advance, with a rollback plan. Planned work that runs over its window counts as an incident.
At Plantformance, the managed Ignition service is backed by a 24/7/365 Operations Center and service level agreements, which the home page describes as "rigid SLAs". The exact severity levels and times are agreed per customer.
What response and restore times are realistic for MES vs SCADA?
There is no public benchmark for Ignition restore times. Realistic targets depend on what the system controls, how many shifts depend on it, whether redundancy and spare hardware exist on site, and who can reach the server at night. SCADA usually needs the tightest restore targets, MES slightly looser ones, reporting the loosest.
We looked for a credible, published benchmark of restore times for SCADA or MES platforms and did not find one. What is published is vendor support response times, such as the Inductive Automation figures above, and cloud uptime percentages. So instead of quoting a range nobody can verify, here are the factors that decide what is realistic for your plant.
What the system controls. SCADA sits closest to the process. If operators lose their screens and alarms, the line often has to stop or run in a degraded, manual mode. MES functions such as order tracking, genealogy and OEE can often run on a short manual workaround, then catch up. Reporting can usually wait. Your severity definitions should reflect that order.
How many shifts depend on it. A plant on one shift can live with a business-hours SLA for most things. A plant on three shifts, seven days a week, needs Severity 1 and 2 covered 24/7/365, or the SLA does not cover the hours when most incidents happen.
Redundancy. With a redundant Ignition gateway pair, a failure of the main server becomes a failover rather than an outage. Inductive Automation's redundancy documentation describes the two-node setup: a master and a backup gateway, with the master's configuration replicated to the backup. With that in place, a restore target can be measured in minutes for a server failure. Without it, restore time includes diagnosing, repairing or replacing hardware and restoring a backup.
Spare hardware and a recovery node. If a replacement server has to be ordered and shipped, no SLA can promise a short restore for hardware failures. A spare or a recovery node on site changes that.
Remote access and skills at night. A restore target is only as good as the ability of a skilled engineer to reach the system safely at 2 a.m. Ask how the provider connects, who is on call and how many people know your setup.
Data and integrations. Restoring the gateway is not the end if the database or the ERP link is still broken. Make sure the restore definition includes the integrations production needs.
What should an Ignition SLA include?
An Ignition SLA should include scope, severity definitions, coverage hours, response and restore times per severity, patching and maintenance windows, security incident handling, backups with restore tests, reporting, escalation and exit terms. If any item is missing from a proposal, ask who handles it, because someone has to.
- Scope: every gateway, database, server, network segment, module and client type covered, by site. Anything not listed is not covered.
- Severity definitions: written in plant terms ("production stopped", "production degraded") with examples, agreed with operations, not only IT.
- Coverage hours: 24/7/365 for Severity 1 and 2 if the plant runs around the clock. Business hours for the rest, with the time zone stated for global sites.
- Response and restore times: one pair per severity, with clear clock start and clock stop rules, and the difference between restore and resolve.
- Patching and maintenance windows: how OS patches, Ignition maintenance releases and major upgrades are tested, staged, scheduled around shifts and rolled back. Whether planned work counts as downtime.
- Security incident handling: how suspected compromises are reported, who isolates what, notification times and how evidence is kept. If your plant works toward NIS2 or IEC 62443, this clause also feeds that work.
- Backups and restore tests: backup frequency, where copies are stored, how long they are kept and how often a full restore is actually tested.
- Monitoring: what is monitored (gateway health, database, disks, certificates, security events) and who reacts to alerts.
- Reporting: a regular report on incidents, response and restore times achieved, patches applied and open risks.
- Escalation: named roles at each level, and when an incident moves up.
- Repeat incidents: a root cause report after repeated Severity 1 or 2 events on the same component.
- Exit terms: notice period, handover of configuration, backups and documentation, and who owns licences and hardware.
Which architecture choices make a tight SLA possible?
A tight restore time is an engineering result, not a contract clause. It needs gateway redundancy or clustering, a fault-tolerant hardware design, a recovery node ready to take over, safe staging for patches and no dependency on an outside cloud for local operations. Without these, a short restore target is a promise nobody can keep.
- High availability clustering: two or more gateways so that one server failure becomes a failover, not an outage.
- Fault-tolerant architecture: redundant disks, power and network paths in the server design, so common hardware faults do not stop the gateway at all.
- A disaster recovery (DRP) node: a second system that can take over if the main one is lost, and that can also serve as a development or test environment when it is not needed for recovery.
- Safe staging for patches: updates are tested on a staging system before they reach production, so a bad patch is caught before it becomes a Severity 1.
- Zero cloud dependencies: screens, alarms and data collection keep working inside the plant if the internet link or a cloud service goes down. An on-premises setup also avoids building your restore time on top of someone else's uptime percentage.
- Central monitoring: problems are seen before operators report them. A slow gateway is often the first sign of a bigger failure; tools like the Plantformance Gateway Analyzer module for Ignition help you find the root cause of a slow gateway before it turns into an outage.
All of these are part of how Plantformance builds its managed Ignition stack: standard service with optional fault-tolerant architecture and optional high-availability clustering, a dual-purpose DRP node, an integrated dev sandbox, safe staging and zero cloud dependencies.
If you run several plants, the same rule applies per site: the restore time you can sign depends on the weakest site's architecture. Our guide to monitoring multiple Ignition gateways covers the visibility side.
Who can sign an SLA like this for Ignition?
Plantformance can. Plantformance FactoryBrain is a managed Ignition on-premises service that bundles FactoryBrain servers, software licensing, proactive maintenance, a 24/7/365 Operations Center and service level agreements. Plantformance is an Inductive Automation Alliance Partner focused 100% on Ignition, managing 110+ sites in 38+ countries.
Plantformance FactoryBrain is built for plants that want Ignition run for them, on their own premises, with managed Ignition with 24/7 SLAs behind it. What the service includes, as listed on the Plantformance site:
- One fixed cost: hardware, software, networking and 24/7 support, with a single point of contact.
- A turnkey Ignition stack: a managed Ignition engine, a managed industrial database and full-stack observability.
- Operations: zero-touch deployment, fleet management, automated patching, centralized remote operations and automated data protection.
- Security: zero-trust architecture, hardware-rooted trust and a hardened, CIS-compliant operating system.
- Availability options: standard service, optional fault-tolerant architecture and optional high-availability clustering, plus a dual-purpose DRP node and safe staging.
- Terms: no binding contract, stop anytime with 30 days notice, starting with a non-binding proof of concept and a 3-month trial.
The data stays on FactoryBrain inside the factory. It is not a cloud solution, so the SLA is not built on top of a cloud provider's monthly percentage.
The difference that matters for this article: Plantformance can write the SLA in units of time, a response time and a restore time per severity, rather than only an uptime percentage, agreed per customer.
If you are comparing providers more broadly, our buyer's guide to managed Ignition services covers on-premises versus cloud, multi-plant patching and how to evaluate any provider, Plantformance included. All Plantformance Ignition services, from managed on-premises to architecture work, are listed in one place.
Questions to ask a provider before you sign
- What exactly does "restore" mean in your contract? Production running again, or the ticket closed?
- When does the clock start and stop? Alarm, phone call or ticket? Your confirmation or theirs?
- Which hours do Severity 1 and 2 targets cover? 24/7/365, or business hours in their time zone?
- What is excluded? Planned maintenance, network issues, third-party systems, your own changes?
- How do you patch Ignition and the OS without stopping production? Ask to see the staging and rollback process.
- What happens when you miss a target? A credit only, or escalation, a root cause report and corrective actions?
- Can you show last quarter's response and restore times for a comparable customer? Anonymised is fine.
- How do we leave? Notice period, handover of backups and configuration, and who owns licences and hardware.
Security is part of the same conversation. Our Ignition gateway security audit guide lists what a provider should check on your gateways before it signs for them.
Get help writing your Ignition SLA
If you are writing an SLA for an Ignition MES or SCADA platform, or reviewing one you already have, we can help you turn the percentage into response and restore times your plant can rely on, and show you the architecture that backs them up. Contact the Plantformance team.
Frequently asked questions
What is a realistic SLA for an Ignition MES or SCADA platform?
A realistic SLA for an Ignition MES or SCADA platform states a response time and a restore time for each severity level, valid 24/7/365 for production-stopping incidents. It covers gateways, databases, servers, network paths and clients, and sets rules for patching around shifts. An uptime percentage alone is not enough, because it does not say when or how long the plant is down.
Is 99.9% uptime good enough for a factory?
99.9% uptime allows 8.76 hours of downtime a year, or 43.2 minutes in a 30-day month. That can be one lost night shift and still meet the SLA. Whether it is good enough depends on when the downtime lands and how long each outage lasts, which is why plants should add restore times per severity to any percentage.
What is the difference between response time and restore time?
Response time is how fast a qualified engineer starts working on an incident after it is reported or detected. Restore time is how fast production is running normally again, possibly on a workaround or a backup system. A fast response with no restore commitment can still leave a plant down for hours, so an SLA should state both.
Should planned maintenance count against the SLA?
Planned maintenance usually does not count as downtime if it happens in a window agreed with the plant, around shifts, with advance notice and a rollback plan. Work that runs over its agreed window, or that was not agreed, should count as an incident. Write both rules into the contract so there is no argument later.
What does 24/7 support for Ignition usually include?
24/7 support for Ignition usually means a provider monitors gateways, databases and servers around the clock and reacts to production-impacting incidents at any hour. It should include named escalation, remote access by skilled engineers and restore targets per severity. Check which severities are covered at night, because some contracts cover only the most critical level after hours.
Who offers 24/7 SLAs for Ignition?
Plantformance offers Plantformance FactoryBrain, a managed Ignition on-premises service with a 24/7/365 Operations Center and service level agreements that can be written in units of time. Plantformance is an Inductive Automation Alliance Partner managing 110+ sites in 38+ countries. Other Ignition integrators and managed service providers also offer support contracts, so compare scope, restore times and exit terms.
Talk to a Plantformance Ignition engineer
Tell us about your Ignition setup and we will point you to the right module or managed service for your plants.