> For the complete documentation index, see [llms.txt](https://docs.seemoredata.io/external-docs/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.seemoredata.io/external-docs/fundamentals/our-features/automations/smart-pulse/burst-protector.md).

# Burst Protector

A safety net for Smart Pulse that automatically handles sudden query spikes by temporarily adding warehouse capacity, in a Cost or Performance mode you choose.

## Overview

Smart Pulse optimizes your Snowflake warehouses by right-sizing them according to a schedule. Downsizing introduces a risk: a sudden spike in query volume can overwhelm the reduced capacity, causing queries to queue up and degrade performance.

**Burst Protector is a safety net** that watches every warehouse running a Smart Pulse schedule and temporarily adds capacity when sustained query queueing is detected. Once the pressure subsides, it scales back down to the scheduled configuration.

Burst Protector runs in one of two modes, chosen per warehouse:

* **Cost** — protection that never costs more than your original, pre-optimization configuration. This is the default.
* **Performance** — protection that may scale *above* your original configuration, up to a ceiling you define, when queueing persists.

{% hint style="info" %}
**Required role to configure:** [Editor](/external-docs/fundamentals/settings/user-roles.md) or higher.

Users with lower roles can view the Burst Protector mode and its activity but cannot change them.
{% endhint %}

{% hint style="info" %}
Burst Protector is always on for every applied Smart Pulse schedule. You do not enable or disable it separately; you only choose how far it is allowed to scale.
{% endhint %}

***

## How to configure

Burst Protector is configured inside the **Smart Pulse Editor**. Open a warehouse details page, go to **Configuration Settings > Schedule**, and create or edit the Smart Pulse schedule. The **Set Burst Protector mode** card sits at the top of the editor.

<figure><img src="https://3620459840-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FnSbIoBjUxWTGNWa9gGw7%2Fuploads%2Fgit-blob-bb913223eb88382ebacd22815085869795455543%2Fburst-protector-modes.png?alt=media" alt="The Set Burst Protector mode card in Cost mode and in Performance mode"><figcaption><p>The mode card in Cost mode (top) and Performance mode (bottom). Performance mode reveals an <strong>Edit settings</strong> button.</p></figcaption></figure>

1. Choose **Cost** or **Performance**.
2. In Performance mode, click **Edit settings** to review the scaling ceiling and timing (see [Performance mode settings](#performance-mode-settings)). Click **Save** in the settings dialog to keep your changes in the editor.
3. Click **Apply schedule**. The mode and settings are saved together with the schedule. Nothing is saved until you apply, and **Cancel** discards all pending changes.

### Default mode

The first time you open the editor for a warehouse, the mode follows the warehouse's Smart Pulse optimization preference: warehouses optimized for **performance** default to Performance mode, and all others default to Cost mode. Once you apply a schedule, your chosen mode is remembered.

### When Performance mode is unavailable

Performance mode adds capacity on top of the warehouse's own configuration. If the warehouse is already at the top of what Burst Protector can scale to (a maximum of 10 clusters on Enterprise editions, or size 6X-Large on Standard edition), the Performance option is disabled and a tooltip explains why. The warehouse continues to run in Cost mode.

***

## Cost mode

Cost mode reverses Smart Pulse's downsizing for as long as a burst lasts, and never goes further. The extra capacity is **capped at the cost of your original, pre-optimization configuration**, so you never pay more than you were paying before Smart Pulse.

Cost mode adapts to your Snowflake edition:

| Edition                                  | Scale-up action                                         | Scale-down action               | Cost cap                                               |
| ---------------------------------------- | ------------------------------------------------------- | ------------------------------- | ------------------------------------------------------ |
| **Enterprise and above** (multi-cluster) | Adds clusters, one at a time, per sustained burst cycle | Removes clusters, one at a time | Total compute never exceeds the original configuration |
| **Standard** (single-cluster)            | Restores the warehouse to its original size             | Returns to the scheduled size   | Never exceeds the original size                        |

For multi-cluster warehouses, the cap accounts for the size ratio between the original and scheduled configuration (for example, Large to Small is 4x) as well as generation differences, so total cluster capacity always stays within the original cost envelope.

***

## Performance mode

Performance mode is for warehouses where query latency matters more than the last dollar of savings. When queueing persists, Burst Protector adds capacity **beyond your original configuration**, up to a ceiling you set. Because this creates spend that no Smart Pulse saving offsets, Performance mode is opt-in and always bounded by your ceiling.

Which lever Burst Protector uses depends on your Snowflake edition:

| Edition                                  | Lever          | Scale-up action                             | Ceiling you set                                                                                           |
| ---------------------------------------- | -------------- | ------------------------------------------- | --------------------------------------------------------------------------------------------------------- |
| **Enterprise and above** (multi-cluster) | Cluster count  | Adds one cluster per sustained burst cycle  | **Add clusters up to** — a maximum cluster count, from one above the warehouse's current maximum up to 10 |
| **Standard** (single-cluster)            | Warehouse size | Moves up one size per sustained burst cycle | **Scale compute size up to** — a maximum size, chosen from the sizes above the warehouse's current size   |

{% hint style="warning" %}
Each cluster adds cost linearly, but each warehouse size step **doubles** the hourly cost. Seemore therefore defaults the ceiling to a single step above the warehouse's own configuration: one extra cluster, or the next size up. Raise the ceiling only if your workload justifies it.
{% endhint %}

### Performance mode settings

Click **Edit settings** on the mode card to open the settings dialog. The first field depends on your edition; the two timing fields are the same for both.

{% tabs %}
{% tab title="Enterprise (multi-cluster)" %}

<figure><img src="https://3620459840-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FnSbIoBjUxWTGNWa9gGw7%2Fuploads%2Fgit-blob-8d1846d48cbc7248d461e7f6fe043731324fd267%2Fburst-protector-settings-enterprise.png?alt=media" alt="Edit Burst Protector settings dialog for an Enterprise warehouse, with the Add clusters up to field"><figcaption><p>Enterprise warehouses set a maximum cluster count.</p></figcaption></figure>
{% endtab %}

{% tab title="Standard (single-cluster)" %}

<figure><img src="https://3620459840-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FnSbIoBjUxWTGNWa9gGw7%2Fuploads%2Fgit-blob-c89d1a02da9c453760d31cb19363ce946f91bb6c%2Fburst-protector-settings-standard.png?alt=media" alt="Edit Burst Protector settings dialog for a Standard warehouse, with the Scale compute size up to field"><figcaption><p>Standard warehouses set a maximum warehouse size.</p></figcaption></figure>
{% endtab %}
{% endtabs %}

| Setting                       | Editions   | What it controls                                                           | Range                                                   | Default                               |
| ----------------------------- | ---------- | -------------------------------------------------------------------------- | ------------------------------------------------------- | ------------------------------------- |
| **Add clusters up to**        | Enterprise | The highest cluster count Burst Protector may reach                        | 2 to 10 clusters, above the warehouse's current maximum | One cluster above the current maximum |
| **Scale compute size up to**  | Standard   | The largest size Burst Protector may reach                                 | Any size above the current size                         | One size above the current size       |
| **Only after queueing lasts** | Both       | How long queries must keep queueing before Burst Protector scales up       | 30 to 3,600 seconds                                     | 180 seconds                           |
| **Return to normal after**    | Both       | How long the queue must stay clear before Burst Protector scales back down | 30 to 3,600 seconds                                     | 180 seconds                           |

Click **Save** to keep the settings in the editor, then **Apply schedule** to put them into effect. If a ceiling can no longer be applied (for example, the warehouse was resized so the ceiling is no longer above its configuration), the apply is stopped and the reason is shown so you can adjust it.

### Guardrails in Performance mode

* **Your ceiling is a hard limit.** Burst Protector never adds more clusters, or moves to a larger size, than you allowed.
* **Every burst episode is time-boxed.** A Performance-mode scale-up lasts at most 30 minutes. When the time is up, the warehouse returns directly to its scheduled configuration, even if queueing continues.
* **Manual changes win.** If someone changes the warehouse's size or cluster settings during a burst, Burst Protector stops managing that episode and leaves the warehouse as it was set.
* **Scale-down is deliberate.** Burst Protector scales back only once the queue has stayed clear for the **Return to normal after** duration and the current load fits the scheduled configuration, so a warehouse does not bounce up and down under one continuous burst.

***

## How it works

Burst Protector samples each running warehouse **once a minute** and compares the number of queued queries with the warehouse's concurrency capacity.

### Burst detection

A burst is detected when **queued queries exceed the warehouse's concurrency capacity**. To avoid reacting to brief spikes, the queue must stay over capacity for a sustained period before Burst Protector acts:

* In **Cost mode**, three consecutive samples must show queueing, so a scale-up follows roughly **three minutes** of sustained pressure.
* In **Performance mode**, the sustain period is your **Only after queueing lasts** setting.

### Scale-up

When a sustained burst is confirmed, the warehouse is scaled up by one step (one cluster, or one size). If queueing persists for another full sustain period, Burst Protector takes another step, until it reaches the cap for the current mode.

### Scale-down

Once no burst is detected for the cooldown period, Burst Protector steps back down toward the scheduled configuration, one step at a time, until the warehouse is fully back to what Smart Pulse scheduled.

***

## Monitoring

### Change log

Burst Protector actions appear in the warehouse change log with a dedicated **"SmartPulse Burst Protector"** badge, so you can distinguish automated burst responses from other changes.

<figure><img src="https://3620459840-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FnSbIoBjUxWTGNWa9gGw7%2Fuploads%2Fgit-blob-1b919a35823025d770a63ad6a06cf5eb11cbfeb0%2Fburst-protector-changelog.png?alt=media" alt=""><figcaption><p>Change log entries showing Burst Protector actions</p></figcaption></figure>

### Real-time stats

Burst Protector activity is also visible in the warehouse's real-time stats view, giving you immediate visibility into active burst responses.

<figure><img src="https://3620459840-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FnSbIoBjUxWTGNWa9gGw7%2Fuploads%2Fgit-blob-38d203da38c4eefbe0fde4fd974d58cdf3938c1f%2Fburst-protector-realtime.png?alt=media" alt=""><figcaption><p>Real-time stats showing the burst protector indicator</p></figcaption></figure>

***

## FAQ

#### Does Burst Protector cost extra?

Burst Protector is included with Smart Pulse. In **Cost mode**, the scaling it performs is always capped at your original warehouse cost, so you never pay more than before Smart Pulse optimization. In **Performance mode**, you explicitly allow spend above your original configuration, bounded by the ceiling you set and the 30-minute limit per burst episode.

#### Which mode should I choose?

Choose **Cost** for warehouses where savings matter most and occasional queueing is acceptable; it is the default and carries no extra cost. Choose **Performance** for latency-sensitive warehouses where a short queue is more expensive to the business than a few extra credits.

#### Why do I see a cluster setting on one warehouse and a size setting on another?

Burst Protector uses the lever your edition provides. Snowflake **Enterprise** editions and above support multi-cluster warehouses, so Burst Protector adds clusters. **Standard** edition has single-cluster warehouses only, so Burst Protector changes the warehouse size instead.

#### Can I turn Burst Protector off?

Not as a separate switch. Every applied Smart Pulse schedule runs with Burst Protector, because it is what keeps downsizing safe. Choosing **Cost** mode keeps protection strictly within your original cost.

#### Can I use Burst Protector without Smart Pulse?

No. Burst Protector requires an active Smart Pulse schedule because it works by temporarily adjusting Smart Pulse's scheduled configuration when bursts are detected.

#### How quickly does Burst Protector react?

In Cost mode, a scale-up follows roughly three minutes of sustained queueing. In Performance mode, you set the sustain period with **Only after queueing lasts**, from 30 seconds to one hour.

#### What if my warehouse is not running?

Burst Protector only monitors warehouses that are currently running. Suspended or idle warehouses are not affected.

#### Can Burst Protector scale my warehouse beyond its original size?

Only in **Performance mode**, and only up to the ceiling you set. In Cost mode, total compute capacity never exceeds the original, pre-optimization configuration.