Skylar Analytics and Advisor are deployed on a Kubernetes infrastructure. The sizings in this chapter below cover the necessary resources for only the applications, not the entirety of the Kubernetes cluster that hosts the application.
Additional resources are required for system and cluster-level services, such as DNS, networking, security. As a result, do not size a node to these requirements alone. ScienceLogic cannot determine the additional overhead needed for cluster components; this must be accounted for separately.
Not all components are currently highly available (HA). Only critical services support HA today, with additional components for HA support coming in future releases.
Node Storage Capacity: Around 350gbs per node of internal storage for images and Kubernetes-related functions. Ensure the root directory has most of this space provisioned. Storage below is referencing Attached storage (EBS).
Sizing for Skylar Analytics
With High Availability
| Deployment | Datapoints per Minute (DPM) Based on Cluster DPM | Pods | CPU | Memory | Storage |
|---|---|---|---|---|---|
|
Micro (POC) |
0 - 5,000 |
35 |
9.66 |
50.9 Gi |
211 Gi |
|
Small |
5,000 - 30,000 |
39 |
21.39 |
100.1 Gi |
766 Gi |
|
Medium |
30,000 - 215,000 |
39 |
39.16 |
271.4 Gi |
976 Gi |
|
Large |
215,000 - 300,000 |
39 |
65.67 |
503.1 Gi |
1.03 Ti |
|
Extra Large |
300,000 - 900,000 |
39 |
111.61 |
879.2 Gi |
5.77 Ti |
Without High Availability
| Deployment | Datapoints per Minute (DPM) | Pods | CPU | Memory | Storage |
|---|---|---|---|---|---|
|
Micro (POC) |
0 - 5,000 |
32 |
8.86 |
50.3 Gi |
207 Gi |
|
Small |
5,000 - 30,000 |
32 |
18.63 |
76.2 Gi |
506 Gi |
|
Medium |
30,000 - 215,000 |
32 |
34.67 |
228.3 Gi |
704 Gi |
|
Large |
215,000 - 300,000 |
32 |
58.30 |
426.6 Gi |
768 Gi |
|
Extra Large |
300,000 - 900,000 |
32 |
101.95 |
795.8 Gi |
4.70 Ti |
Additional Sizing for Adding Skylar Advisor to Skylar Analytics
Add the following resources to the existing provisioned Skylar Analytics resources.
The storage numbers are for backend services and include 20 GB reserved for the Skylar Advisor Corpus. The Corpus can be much larger, so you must account for additional storage specifically for the Corpus.
With High Availability
| GPU Profile | + Pods | + CPU | + Memory | GPU | + Storage |
|---|---|---|---|---|---|
|
4 GPU (l40s/RTX) |
32 |
74.52 |
407.8 Gi |
4 |
394 Gi |
|
8 GPU (h100) |
38 |
99.82 |
532.0 Gi |
8 |
394 Gi |
Without High Availability
| GPU Profile | + Pods | + CPU | + Memory | GPU | + Storage |
|---|---|---|---|---|---|
|
l40s-rtx6000-4gpu |
30 |
69.92 |
371.0 Gi |
4 |
394 Gi |
|
h100-resources |
36 |
95.22 |
495.2 Gi |
8 |
394 Gi |
Largest Single Pod Sets Minimum Node Size
At least one node must be large enough to accommodate the ClickHouse database. This requirement is generally not an issue for smaller deployments, but it becomes an important consideration for Medium deployments and larger.
For example, a Medium deployment requires a ClickHouse node with at least 18.52 cores and 147.4 Gi of memory. After you set aside the ClickHouse node, you can size the remaining nodes evenly.
| Deployment | Largest Pod | CPU | Memory |
|---|---|---|---|
|
Micro |
n/a |
n/a |
n/a |
|
Small |
ClickHouse |
4.71 |
18.5 Gi |
|
Medium |
ClickHouse |
18.52 |
147.4 Gi |
|
Large |
ClickHouse |
36.91 |
294.5 Gi |
|
Extra Large |
ClickHouse |
73.72 |
588.9 Gi |
Data Visualization (DataViz) Sizing
The previous sizing tables are for the Skylar AI application. Skylar AI can be multi-tenant, and the sizing recommendations change as you add tenants. Add the following requirements on top of the requirements in the previous tables, per tenant.
| Tenant Size (Based on Tenant DPM) | DataViz Resourcing |
|---|---|
|
Micro |
5 CPU and 32 GB RAM |
|
Small |
5 CPU and 32 GB RAM |
|
Medium |
5 CPU and 64 GB RAM |
|
Large |
5 CPU and 64 GB RAM |
|
Extra Large |
8 CPU and 128 GB RAM |
Additional Sizing for Adding Streaming Telemetry
If you are including Streaming Telemetry, add the following resources to the existing provisioned Skylar Analytics resources. The requirements are the same for every deployment size, with and without HA.
| CPU | Memory | Storage |
|---|---|---|
|
1.67 |
2.7 Gi |
None |