You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Copy file name to clipboardExpand all lines: content/learning-paths/automotive/tinkerblox_ultraedge/background.md
+84-40Lines changed: 84 additions & 40 deletions
Display the source diff
Display the rich diff
Original file line number
Diff line number
Diff line change
@@ -6,50 +6,94 @@ weight: 2
6
6
layout: "learningpathall"
7
7
---
8
8
9
-
{{% notice Note %}}
10
-
REMOVE ME: Need to review content for Intro/background...
11
-
{{% /notice %}}
12
-
13
-
### Overview
14
-
15
-
UltraEdge was built with the vision of orchestrating the edge-native
16
-
execution fabric for high-performance compute infrastructure
17
-
18
-
- UltraEdge is a ‘built-for-edge’ adaptive **AI & Mixed Workloads**
19
-
execution stack built on the ethos of high performance, high
20
-
fungibility & ultra-low footprint
21
-
- Developed through strategic alliances with world-renowned technology
22
-
powerhouses
23
-
- Clear dual focus on Mixed workloads and new-age AI workloads
24
-
- Full stack enablement through MicroStack & NeuroStack systems
25
-
- Curated for AI@Edge with preferred edge deployment approach by Edge
26
-
AI Foundation
27
-
- Managed cluster” orchestration through integration with Kube-stack
28
-
and/or Slurm
29
-
- Observability for control plane, diagnostics & telemetry
30
-
- Demonstrable value to customer through lower TCO of CPU-GPU clusters
9
+
### Introduction
10
+
11
+
UltraEdge is an edge-native, high-performance execution fabric designed to run AI and mixed workloads without the overhead of traditional container platforms. While technologies like Docker and Kubernetes were created for general-purpose cloud environments, they introduce latency, resource bloat, and non-deterministic behavior that are poorly suited for edge deployments.
12
+
13
+
UltraEdge takes a fundamentally different approach. It replaces heavyweight container runtimes with a lean, deterministic execution stack purpose-built for performance-oriented compute. This enables millisecond-level startup times, predictable performance, and a dramatically smaller resource footprint—allowing workloads to start faster, run closer to the hardware, and make full use of available CPU and GPU resources.
14
+
15
+
At the core of UltraEdge are two specialized execution systems:
16
+
17
+
· **MicroStack**, optimized for enterprise and mixed workloads
18
+
19
+
· **NeuroStack**, purpose-built for AI inference and accelerated compute
20
+
21
+
Together, these systems deliver up to **30× faster startup times** and **3.8× smaller package sizes** compared to conventional container-based approaches. By removing unnecessary abstraction layers, UltraEdge ensures compute cycles are spent on execution—not on managing the runtime itself.
22
+
23
+
This learning path introduces the architecture, principles, and components that make UltraEdge a high-performance execution fabric for modern edge infrastructure.
24
+
25
+
### Ultraedge Overview
26
+
27
+
UltraEdge was built with the vision of **orchestrating the edge-native execution fabric for high-performance compute infrastructure**
28
+
29
+
Key design principles and capabilities include:
30
+
31
+
· **Built-for-edge execution stack**
32
+
33
+
A lightweight, adaptive platform for **AI and mixed workloads** optimized for low latency, high determinism, and minimal footprint.
34
+
35
+
· **Dual workload focus**
36
+
37
+
Native support for both traditional enterprise workloads and next-generation AI workloads, without compromising performance.
38
+
39
+
· **Full-stack enablement**
40
+
41
+
Delivered through the **MicroStack** and **NeuroStack** execution systems, each optimized for its workload domain.
42
+
43
+
· **High fungibility and efficiency**
44
+
45
+
Maximizes utilization of CPU and GPU resources while reducing operational and infrastructure overhead.
46
+
47
+
· **Ecosystem-aligned development**
48
+
49
+
Developed through strategic alliances with leading technology partners and curated for **AI@Edge**, including alignment with Edge AI Foundation deployment approaches.
50
+
51
+
· **Cluster-aware orchestration**
52
+
53
+
Integrates with Kubernetes-based stacks and Slurm for managed cluster orchestration.
54
+
55
+
· **Built-in observability**
56
+
57
+
Provides control-plane visibility, diagnostics, and telemetry for operational insight.
58
+
59
+
· **Lower total cost of ownership (TCO)**
60
+
61
+
Demonstrable reduction in CPU/GPU cluster costs through faster startup, higher utilization, and reduced runtime overhead.
31
62
32
63
### UltraEdge High-Level Architecture
33
64
34
-
{{% notice Note %}}
35
-
REMOVE ME: It would be good to put a high-level picture of the architecture here. Then text below can detail the high points.
36
-
{{% /notice %}}
65
+
UltraEdge is composed of layered systems, each responsible for a distinct aspect of execution and orchestration:
You should be able to locate and copy the specific installation command appropriate for your account. Here is an example:
45
+
46
+
```bash
47
+
sudo DEVICE_ID="5b3ff290-0c88-4cd9-8ef7-08de0bded9df" KEY="TB.ApiKey-mlBZgDFc7qyM6ztPjILBCbFEqnVlbvjUpM1Q1IqNP6tA7wNdi97AQ==" sh -c "$(curl "https://tinkerbloxdev.blob.core.windows.net:443/tinkerbloxdev/binaries/installer.sh?sv=2025-01-05&st=2025-11-03T06%3A31%3A55Z&se=2025-11-03T06%3A56%3A55Z&sr=b&sp=r&sig=HNS70HgJyHlhCVQrqvpGdCcaf8%2FtVjdW4RNiiiIPCSUA%3D")"
48
+
```
49
+
50
+
Run your specific installer command in your Ubuntu/Debian SSH shell and initialize the agent to install UltraEdge.
51
+
52
+
### Activation of the UltraEdge Agent
35
53
36
54
On the first boot, the agent will automatically generate a file named
37
-
`activation_key.json` at the path:
55
+
`activation_key.json` at the path:
38
56
39
-
/opt/tinkerblox/activation_key.json
57
+
`/opt/tinkerblox/activation_key.json`
40
58
41
59
Share this `activation_key.json` file with the TinkerBlox team to
42
60
receive license key (which includes license metadata).
43
61
44
62
1. Stop the agent using the following command:
45
63
46
-
sudo systemctl stop ultraedge.service
64
+
```bash
65
+
sudo systemctl stop tbx-agent.service
66
+
```
47
67
48
68
2. Replace the existing `activation_key.json` file in
49
69
`/opt/tinkerblox/` with the licensed one provided by TinkerBlox.
50
70
51
71
3. Start the agent:
52
72
53
-
sudo systemctl start ultraedge.service
73
+
```bash
74
+
sudo systemctl start tbx-agent.service
75
+
```
54
76
55
-
####Manual Running
77
+
### Manual Running
56
78
57
-
- Binary path: `/usr/bin/EdgeBloXagent`
79
+
- Binary path: `/bin/tbx-agent`
58
80
59
81
- To start:
60
82
61
-
EdgeBloXagent
83
+
```bash
84
+
tbx-agent
85
+
```
62
86
63
87
- To stop, press <spanclass="kbd">Ctrl</span> +
64
88
<spanclass="kbd">C</span> once.
65
89
66
90
## MicroPac Installation
67
91
68
-
{{% notice Note %}}
69
-
REMOVE ME: Is MicroPac only for Debian installations? Not for YOCTO ones?
70
-
{{% /notice %}}
92
+
MicroPac is the core tooling used to build and manage **MicroStack** (general microservices) and **NeuroStack** (AI-native services).
93
+
94
+
***Platform Agnostic:** MicroPac is not restricted to a specific operating system; it is fully compatible with both **Debian** and **Yocto** environments, providing a consistent execution layer across different Linux distributions.
95
+
96
+
***Build System:** To create a service, the system utilizes a **MicroPacFile** (the declarative configuration) and the **MicroPac Builder** (the high-performance packaging engine).
97
+
98
+
***Validation:** The ecosystem includes a **MicroPac Validator**, which verifies the integrity and security of the package created by the builder to ensure it is ready for edge deployment.
71
99
72
100
#### System Requirements
73
101
@@ -78,34 +106,57 @@ REMOVE ME: Is MicroPac only for Debian installations? Not for YOCTO ones?
78
106
79
107
#### Required Packages
80
108
109
+
```bash
81
110
sudo apt-get update
82
111
sudo apt-get install -y tar curl qemu-user-static binfmt-support
112
+
```
83
113
84
114
### Cross-Architecture Support
85
115
86
-
{{% notice Note %}}
87
-
REMOVE ME: Might need a bit more detail on why this needs to be executed (below):
88
-
{{% /notice %}}
116
+
The **MicroPacFile** is the central declarative configuration used by the builder to define the environment and behavior of your service. This configuration is essential for orchestrating both **MicroStack** (general microservices) and **NeuroStack** (AI/ML) services.
117
+
118
+
***Multi-Language Support:** You can configure MicroPacFiles for applications written in **Python, C, and C++**, making it highly versatile for both high-level AI workloads and low-level embedded system tasks.
119
+
120
+
***Unified Workloads:** It bridges the gap between complex ML models and resource-constrained embedded software, ensuring consistent execution across diverse hardware.
121
+
89
122
90
123
To build MicroPac for different architectures:
91
124
# Enable binfmt for armv7
125
+
126
+
```bash
92
127
sudo update-binfmts --enable qemu-armv7
128
+
```
93
129
94
130
### Installation
95
131
96
132
- The package is provided as a `.deb` file.
97
133
98
134
- Install it on your host machine:
99
135
136
+
```bash
100
137
sudo apt install ./<package_name>.deb
138
+
```
101
139
102
140
### MicroPac File Schema file creation/setup
103
141
104
-
{{% notice Note %}}
105
-
REMOVE ME: Need more information on how to setup your project directory/where its located
106
-
{{% /notice %}}
142
+
#### File Placement
143
+
For the MicroPac Builder to function correctly, the **MicroPacFile** must be placed in the root directory alongside your source code and dependency files.
0 commit comments