Every alert, every dashboard and every meshIQ answer in Innomesh starts in the same place: Elastic. Innomate built Innomesh on Elastic from the beginning, because AV only becomes manageable when you can see every device in every room, all the time, and keep that data long enough to learn from it. Elastic is the backbone that makes that possible across more than 100,000 devices.
Elastic has featured Innomesh in a customer story (opens in a new tab). In this post we look at how Elastic works inside Innomesh, and what it means for the teams who run AV estates on it.
Why does AV need a data platform at all?
Walk into a typical university estate and count the places AV data lives. The DSP keeps its own logs. The displays report to one vendor tool, the cameras to another, the control processors to a third. Each sees its own devices and nothing else.
Most of that data disappears before anyone reads it. A Q-SYS core, for example, keeps only its last 5,000 log entries, which on a busy core can be a few days or less. The logs from a fault on Friday can be gone before anyone looks the following week.
So the fault is found the old way. A lecturer walks into a room, the screen is dark, and someone calls the help desk. Your team learns about the problem at the moment it hurts most.
Fixing that needs more than another dashboard. It needs one place where every reading from every device lands, stays, and can be searched.
Why did we build Innomesh on Elastic?
We wanted the data layer to be the strongest part of the platform, because everything we planned to build would sit on it. Elastic gave us four things in one place.
Elasticsearch
One store for every kind of AV data.
Time-series telemetry, device state, logs and room configuration all live in Elasticsearch. Your monitoring, your dashboards and your AI all read from the same source, so they never disagree.
Scale
Speed at estate scale.
A single large tenant writes around 29 million records a day into Elastic. Elastic keeps search and reporting fast as the data from rooms and devices grows.
Machine learning
Machine learning built in.
Elastic's anomaly detection watches for patterns no single alert reveals, such as a burst of alerts across five or more rooms in the same window.
Elastic Cloud
A managed cloud service.
Innomesh runs on Elastic Cloud, so our engineers spend their time building the platform instead of managing data clusters. That fits well with a platform that is predominantly serverless from end to end.
How does Innomesh turn device data into action?
The path from a device to your team has four steps, and Elastic sits at the centre of all of them.
- Step 1
Collect
Workers, native manufacturer connectors, middleware applications, platform-to-platform integrations and log streams.
- Step 2 · Elastic
Store
Device telemetry, state, logs, events and configurations land in Elasticsearch as searchable history, and as a live view of every room.
- Step 3
Detect
Alert rules and machine learning across the whole estate, catching anything from a single device failure to a site-wide outage.
- Step 4
Act
Alerts with AI root cause analysis, device automation and self-healing, and automated security hardening, all available through meshIQ, the Innomesh AI agent.
Collect. Innomesh workers collect data from devices over the network. Innomesh middleware applications also run natively on manufacturer controllers, turning each controller into a managed edge device that reports straight into the platform. Connectors pull from platforms such as Biamp SageVue, Logitech Sync and Appspace. Coverage spans 100+ manufacturers, so a mixed estate reports into one place.
Store. All of that data, from telemetry and state to logs, events and configurations, lands in Elasticsearch, where it becomes searchable history for the whole estate. Innomesh maps raw device data into room state (power, occupancy, source routing), so you get the big picture quickly, whether you're looking at one room, one site or your entire estate.
Detect. Alert rules and machine learning run across the whole estate. A single failing projector raises one alert. A building of disconnected devices raises an incident. The platform adapts how it presents faults to you, in the way that best helps you understand and respond.
Act. Alerts reach your team with AI root cause analysis, sync both ways with ITSM platforms such as ServiceNow and TeamDynamix, and roll up into trackable incidents. Many alerts. One trackable incident.
What does the same data give meshIQ?
An AI layer is only as useful as the data under it. meshIQ, the AI inside Innomesh, draws on the telemetry your estate has already written into Elastic. You talk to your data and it responds to you in dashboards.
Show me the user interaction journey for a typical room over the last 30 days
User interaction journey · busiest room · 30 days
Here is the interaction journey for a typical room over the last 30 days. Source switching and volume adjustments are the top flows.
Analyse why the touch panel has been failing in Room 789.
Confirm action expires in 9:52
- Action
- Analyse Logs (downloads the device log, then analyses the last 7 days)
- Device
- Crestron TSW-1070 · Touch Panel
- Room
- Room 789
ConfirmCancel
Analyse Logs · Touch Panel · Room 789 · 3 findings
Last 7 days on the touch panel. Nothing ties the failures to one cause yet; these are the leads:
- New1,358 kernel watchdog warnings; 3 of 20 panels show this. Fits a service restarting repeatedly.
- CrashHydrogenManager crashed 6 Oct 15:59 (SIGBUS); also 11 Sep. Not yet linked to the failures.
- WarnVideo stream client sometimes sees its own address as 0.0.0.0; steady since at least 25 Sep.
watchdog: watchdog0: watchdog did not stop!
All answers come from one place. Telemetry, logs, live states and history are stored and correlated in Elastic, and Innomesh connects the dots to present just the information you need, in a human-friendly way you can act on.
meshIQ can also fix what it finds. From the chat, you can reboot a device, schedule an action or set up an automation policy. Nothing runs until you confirm it, and meshIQ can only do what you allow it to do, based on the permissions you set.
What changes for your team?
| Without a data plane | With Innomesh on Elastic |
|---|---|
| Faults are reported by the people they affect | Faults are detected and alerted as they happen, wherever they happen |
| Logs roll over in days on each device | Logs are kept for as long as needed, searchable estate-wide |
| Every manufacturer has its own reporting tool | 100+ manufacturers report into one view |
| Patterns across rooms go unnoticed | Machine learning flags patterns across rooms and buildings |
| Answers need someone to decipher the data and build a report | meshIQ answers in plain language, from your own data |
| Support tickets pile up as people report faults | Considerably fewer support tickets, because you know what is happening in every room before anyone tells you |