Release Summary

recommendations with Photon-aware logic
enhanced flagging of All-Purpose Compute-on-jobs recommendations
redesigned recommendations view
adds forecasted DBX/Cloud KPIs to Spend Insights & Forecasting
adds read-only access to Incident Policies for non-admins and
renames Setup Time Dominates Execution Time to Serverless Opportunity
introduces a SQL Warehouse configuration timeline;

Top Recommendations

Photon-aware scale-down with Photon-specific wording and a memory-over/under disclaimer.

All-Purpose Compute-with-jobs recommendations now generate even when the All-Purpose Compute has none of its own, with serverless suggested in the texts.
- Two-way navigation between jobs and All-Purpose Compute — open a job to find the All-Purpose Compute it runs on, or open an All-Purpose Compute to find every job using it as compute.
- All-Purpose Compute clusters used to run jobs now get their own recommendations and incidents, even when the All-Purpose Compute has none of its own, with serverless or job clusters suggested as the fix.

Level 4 recommendations view redesign — refreshed header, banner, tags, table styling, plus a tooltip for cluster-instance cost.

Recommendations main page improvements — top widget shows total spend of recommended workloads, compute-type chips added, "Workload ID" rename + CSV export, gear-icon column control.

Performance — Top Recommendations service faster, level-1 latency reduced, gateway-timeouts on large run sets fixed.

Spend Insights & Forecasting

Forecasted DBX and Cloud KPIs added to the Forecasting page.

Databricks unit plan management — Databricks unit and Price columns on the Databricks Cost Plan table, refreshed plan info / popovers / tooltips, plan period end-date linked to usageEndDate.
- The plan table gained DBCU (committed amount) and Price columns, so plan size and cost are visible at a glance instead of just discount and dates.
- Databricks Cost Plan — the table of pre-purchased Databricks commitment plans now shows each plan's committed DBCU amount and price, so plan size, cost, term and discount are visible in one place.
- Cost plan accuracy — refreshed plan info, popovers and tooltips, with a corrected plan period end-date and per-plan status.

Weekly and monthly polish — selected-week marker, line-indicator color matches legend, projected-cost mark-line intensity, current-month forecast bar fix.
Various forecasting and chart fixes (Last-6-Months date range, Daily-view edge cases, weekly forecast tooltips, Export-to-CSV modal).

Workflows & Pipelines

Unified detection for job runs and clusters to improve performance.
“Foreach” data usage added to workflow cost analysis.
Workflow Jobs table — Data Usage column and an actions menu added.

📸 Screenshot: Workflow Jobs table with the new Data Usage column and actions menu.

One-time runs — surfaced in the All-Purpose Compute view with cluster/appId persistence across tabs.
Pipelines — data-usage breakdown on Pipeline Updates, unique URLs per Pipelines sub-tab, pagination and tooltip fixes.

Serverless Opportunities

What it is. A Serverless Opportunity incident flags a job that spends a large share of its total runtime on cluster startup rather than on real work. The job is paying — in both money and wall-clock time — mostly to spin compute up.

Why it exists. A classic Databricks cluster takes minutes to start. A short job can easily spend more time starting its cluster than actually running. That startup time is billed and adds latency, yet produces nothing. The old name, "Setup Time Dominates Execution Time," described that symptom. The new name, "Serverless Opportunity," names the fix: move the job to serverless compute, which starts almost instantly and removes the overhead entirely.

How to use them. Treat these incidents as a ready-made, prioritized backlog of low-risk serverless migration wins — the jobs where the payoff is clearest. The usual culprits are short, frequently-scheduled jobs (a 30-second job that takes four minutes to start). Migrating each one cuts both cost and end-to-end runtime, with little risk because the job logic doesn't change.

Incidents & Policies

Setup Time Dominates renamed to Serverless Opportunity; its email notification is now off by default.

Read-only access to Incident Policies for non-admins, with a "View" option added.
Bulk subscription flow — bulk subscribe/unsubscribe across all incidents, add an email group to all policies at once, "Manage Notifications in Bulk" relabel.

Cost Control polish — clearer date-range wording, refreshed page layout, "Open in Databricks" added on supported incidents, "view policy" link embedded in emails.
Incident type / subtype counts now consistent when navigating from policies (the +N badge no longer mismatches).
Provisioning incident look-and-feel — over/under-provisioning card refresh and chip-fits-content fix.

Cost Over Time & Workspace Cost

Workspace Cost actions menu — kebab on unpublished workspaces with View Cost Over Time / Open.

Cost-type filter on the Cost Over Time Feature tab.

Workspace Cost design — table redesign for screen space, summary section, refreshed cost columns.
AI / Databricks Platforms supported in optimization review and cost-over-time navigation.
Various Cost Over Time fixes (selected filters persist across tabs, SQL Warehouse stack, Y-axis halfway when total < $1).

Consumption & Billing

vendors-usage pre-aggregated into daily summary tables for fast service reads.
Cost-per-Workspace-Day (CPWD) classification pipeline with EA-discount per-workspace-day table and a refactored the cost classifier.
Azure VM Managed Identity for consumption and storage authentication.
AWS Cost & Usage Report loader hardened against missing-column and column-type-mismatch errors.
Savings Plan pricing model estimation for Azure consumption.
Discounted-rates service endpoints accessible to non-admin users; AWS and Azure instance rates updated.

Provisioning & Deployment

Discovered Active Jobs (Discovered Active Jobs) on Delta — Azure SQL replaced with Delta tables.
Workloads Provisioning design refresh and Monitor Setup filter persistence on refresh.

SQL Warehouses

SQL Warehouse configuration timeline — new service and interface showing configuration changes over time.

Configuration change count added to the report service and as a Change column on the SQL Warehouses list.
List polish — Auto Stop visible without scrolling, column order updated, thousands separator on query count, null Active time resolves to N/A.

Vendor Consolidation

Subscription / Workspace filter supported in Vendor Consolidation (interface + service) and passed to the available-vendors endpoint.
AS/400 added as a supported vendor.
Migration tab UX — filter URL sync, persistent filters on refresh, CSV export honors pageSize, banner hidden when no dependent vendor data.

Serverless

Jobs and One-Time Runs tabs removed from the Serverless view; both now live in the unified workflows view.
ForEach task serverless detection and cluster handling fixed.
Serverless Notebooks hide the Hourly Rate column by default.
Performance — query optimization and indexes for serverless-job telemetry costs.

Interface, Charts and Cross-cutting

Trends-view tooltip with pinning and preview — new interaction model for trends tooltips.

User Activity tab exposed in navigation, with redesigned page and CSV export.

Optimization Review – Jobs (renamed from beta) with refreshed colors and badges.

Subscription / workspace dropdowns — "Clear all", scope-change clears, no resort while selecting, smarter all-vs-one labels.

Cross-page filter consistency — date range preserved across timezones, shared filters synced across Incidents tabs, search filter no longer leaks across workloads tabs.

3.3 Release Notes