mbg@portfolio:~/context-aware · case study 05 · read 6 min
Quiet search --:--:--
$ cat case_05/context-aware.md
// interfaces that adapt to who's looking, what the machine is doing, and what you need

Context-aware enterprise UX.

This didn't come from a PdM roadmap. It came from listening. Sitting with users, watching them manage server hardware remotely, seeing them squint at tables and reconstruct the machine in their heads. iDRAC is the 1:1 hardware-to-software bridge, so we rebuilt the bridge. Dynamically-populated SVG renderings of the actual hardware, with context-aware overlays on top. Seeing at a glance that your 1U has drives in slots 1-7 and 11-13, which drives are in a RAID together after a rebuild, and which one is the hot spare. The guesswork, gone. Shipped in iDRAC. The patents came after.

Role
Inventor & UX engineer
Concept through shipped product
Timeline
2016 → Present
Across PowerEdge 14G — 17G
Surface
iDRAC
Integrated Dell Remote Access Controller
Patents
3+ disclosures
Device Imagery · Capacity Overlay · Bay Capability
Device Imagery concept - iDRAC storage view with front, rear, and mid chassis SVG renderings and hover detail showing Hard Drive 2:17 status
FIG. 00 Device Imagery. Dynamically populated SVG renderings of the actual chassis (front, rear, mid) with real-time state per bay and per drive. Hardware as its own visual language, rather than parsed out of a table.
01 The Problem context

Hundreds of components. Text-heavy tables. A disconnect between the UI and the machine in front of you.

iDRAC and every comparable product represented sub-component state as tables. Drives, bays, sensors, RAID groups, virtual disks, hot spares, capacity, thermal, firmware. Hundreds of rows per server. Operators had to parse them, hold the chassis in their head, and cross-reference one against the other. The most expensive version of this was capacity planning, and in passthrough mode operators often had to leave iDRAC entirely for a third-party UI to see drive fullness.

Problem overview - current system management restricts visibility to text-heavy tables
FIG. 01Text-heavy tables of virtual disks, RAID groups, and physical disks. Plausibly complete, practically unusable at scale.
Problem statement - capacity planning requires understanding free or remaining storage at drive, slot, and volume level
FIG. 02Capacity planning specifically. Users couldn't tell which drive in which slot was approaching fullness without leaving the tool.
02 The Insight concept

Let the hardware be its own visual language.

Stop translating the machine into a table. Render it. Every Dell chassis has a known physical topology and every deployed system has a known configuration, which makes a chassis drawable. Device Imagery is the pattern that did the drawing, a pre-authored SVG library of every chassis Dell ships, paired with a discovery layer that populates it from real-time system state. Once the base layer existed, the interesting question was what to draw on top of it.

03 The Pattern Family 3 overlays · 1 base

One visual base. Multiple context-aware overlays. Each one reads the machine and the task.

Device Imagery was the substrate. The real UX work was in the overlay system, a family of context-aware layers that render on top of the same SVG, each surfacing a different dimension of state for a different operator task. Two are worth walking through, each covered by its own filing in the patent family.

O.01 Capacity-View Overlay.

Answers what the tables couldn't: where, physically, is capacity being used? Colors every drive slot by fill state, scoped to what the operator has selected (all drives, a RAID group, or remaining space after selected VDs). Passthrough mode is handled natively, no third-party UI required.

Flow chart for capacity-view overlay system with shipped UI screenshots showing All Drives, RAID Group, and Virtual Disk capacity views
FIG. 03Capacity-view overlay. Decision flow on the left, shipped behaviors on the right.

O.02 Bay Capability Overlay.

Answers a different question, one that got more important as CXL entered the portfolio: what is each bay capable of, and how do I upgrade? Hover a bay to see slot count, bus, media-type support, CPU affinity, backplane version, CXL capability. If a bay can upgrade to CXL, the action surfaces inline. If it can't, the overlay says so plainly.

Bay Capability overlay flow chart showing CXL detection and upgrade path with shipped UI screenshots
FIG. 04Bay Capability overlay. CXL-capable bays surface upgrade inline, non-CXL bays say so plainly.
A note on provenance

Device Imagery, the capacity-view overlay, and the bay-capability overlay are each covered by separate filings. The pattern family is part of the 33 filed / 23 granted body of work from serving on the Enterprise HW Patent Committee. Specific claims and filings available on request.

04 Outcome shipped

Shipped in iDRAC. The pattern became the default.

Device Imagery and its overlay family shipped in iDRAC as Bay View. The SVG library and discovery layer now get reused across chassis SKUs. Operators don't describe this as "adaptive" or "context-aware." They describe it as "I can see the server." A context-aware UI that announces its context-awareness has failed.

Shipped iDRAC dashboard with Device Imagery integration
FIG. 05Shipped. Device Imagery integrated into the iDRAC dashboard as Bay View.
→ NEXT_CASE

Synthetic Enterprise Personas · research-grounded scale.

Continue → 06/personas