Siten Sanghvi / Granted July 21, 2020
Smart sensors that watch every node — and call in help when they can't fix it themselves.
An edge computing system that attaches SNMP-driven smart sensors to every data center component, detects configuration changes, classifies intent, and routes peer-to-peer remediation — without a central controller.
US10721135B1Patent number
2019-06-17Filed
1 yr 1 moTime to grant
16 Claims2 independent
26 CitationsForward citations
SCROLL TO EXPLORE
Visual patent explainer
01 / The operational pain
Data centers that can't diagnose themselves.
SNMP was built to let administrators observe a network — not to let the network fix itself. When something changes, a human has to notice, triage, and act.
Meanwhile, disaster recovery depended on reaching a geographically separate center — expensive, slow, and often unavailable at the moment of failure.
Three gaps this patent addresses
01
No autonomous responseSNMP gives visibility, but a change on a node stays broken until a human intervenes.
PASSIVE02
No peer-to-peer repairA node that can't self-heal has no mechanism to recruit a neighbor that can.
ISOLATED03
Remote-only disaster recoveryReaching an off-site DR center adds latency, cost, and a single point of dependency.
REMOTE02 / The inventive move
Give every node a microprocessor and a conscience.
The core idea: attach a smart sensor — an embedded sensor, a low-power microprocessor, and an SNMP agent — to every data center node. When something changes, that node investigates.
If the change was intentional, it passes through. If not, the node tries to reverse it. And if it can't, it finds another node in the network that can — and hands the task off.
The three-step response loop
01 / DETECTSNMP agent on the smart sensor detects a change in a physical or configuration attribute.
02 / CLASSIFYMicroprocessor investigates cause. Was the change intentional or unintentional?
03 / REMEDIATESelf-heal if possible. If not, identify a peer node that can reverse the change and transmit the fix.
03 / System architecture
A mesh of sensing, deciding, and fixing.
The system is a two-layer mesh: data center nodes (the hardware infrastructure) and smart sensors (the intelligence layer) — each sensor permanently affixed to its node and communicating via SNMP.
When a first node cannot remediate a change, it transmits change data directly to the smart sensor of a second node — bypassing any central management plane.
Simplified system topology
DATA CENTER NODE AAny infrastructure component: server, switch, router, plug, outlet, camera…
SMART SENSOR AEmbedded sensor + microprocessor + SNMP agent. Always attached to Node A.
SNMP AGENT → NODE BAgent-to-agent transmission: Node A sends change data to Node B's sensor when it cannot self-heal.
SMART SENSOR BReceives change data, executes the reverse on behalf of Node A.
In multi-datacenter configurations, the patent treats each data center as a node — the same peer-remediation protocol applies at datacenter scale.
04 / The sensor layer
One sensor. Three components. Every node.
Each smart sensor is a self-contained intelligence unit. It senses changes at the physical layer, processes them locally on a low-power microprocessor, and communicates via SNMP — consuming minimal resources while staying always-on.
The patent lists the full scope of eligible nodes: servers, switches, routers, electrical plugs, outlets, power supplies, displays, modems, printers, IP phones, IP cameras, computer hosts, and combinations thereof.
Smart sensor anatomy
EMBEDDED SENSORDetects changes in physical operation attributes (power state, switch status, light state) and configuration attributes (settings, parameters).
LOW-POWER MICROPROCESSORLocally processes detected changes — classifies intent, attempts remediation, and identifies peer nodes. Always-on with minimal energy overhead.
SNMP SOFTWARE AGENTStandard network management protocol layer. Handles detection, alert routing, and agent-to-agent communication with other smart sensors.
COMMUNICATION CAPABILITIESWired or wireless. Claimed variants include NFC (Near Field Communication) and Bluetooth — enabling close-range peer coordination without network infrastructure.
05 / The decision gate
Not every change needs fixing.
The most important design choice in the system: before attempting any remediation, the microprocessor classifies the change. Intentional changes — from authorized admins or planned updates — are allowed through untouched.
Only unintentional changes trigger the remediation path. This gate prevents the system from fighting legitimate configuration work and allows self-healing to coexist with normal administration.
Intent classification logic
⚡
SNMP AGENT DETECTS CHANGEPhysical operation attribute (power, switch state, light) or configuration attribute changes on a node.
?
MICROPROCESSOR INVESTIGATESDetermines whether the change was intentional (authorized, planned) or unintentional (failure, unauthorized, drift).
✓
INTENTIONAL → PASS THROUGHNo action. Legitimate administration is never blocked or reversed.
ALLOW
↺
UNINTENTIONAL → REMEDIATIONAttempt self-reversal first. If unable, escalate to peer node protocol.
ACT
06 / The peer protocol
When self-repair fails, find a neighbor.
This is the novel peer-to-peer layer. If a node's own sensor cannot reverse a change, the system doesn't wait for human intervention or escalate to a central controller. It scans the network for a node that has the capability to perform the reversal.
Change data travels directly between smart sensors via the SNMP agent layer. The second node executes the fix on behalf of the first — a distributed, autonomous remediation mesh.
Peer remediation protocol / step-by-step
1
SELF-REPAIR ATTEMPTMicroprocessor on Node A's sensor tries to reverse the unintentional change directly on Node A.
2
IDENTIFY CAPABLE PEERIf unable, microprocessor scans the plurality of data center nodes to find one with the capability to reverse the change.
3
TRANSMIT CHANGE DATANode A's SNMP agent sends change data to the smart sensor of the identified Node B.
4
PEER EXECUTES REVERSALNode B's sensor executes the fix on behalf of Node A. No central management plane required.
07 / Proactive capacity management
Alert before the threshold breaks.
Beyond change detection, the system monitors resource utilization. When a server nears its capacity threshold, the sensor transmits an alert — and can optionally place an automatic order to upgrade resources.
The system also detects duplicate data records, copies them, and eliminates the need to reach a geographically separate disaster recovery center — local redundancy replaces remote dependency.
Resource threshold monitoring / three scenarios
SERVER ALPHA — COMPUTE
41% utilization — nominal. No alert. Sensor monitors passively.
SERVER BETA — MEMORY
78% utilization — nearing threshold. Alert transmitted. Auto-upgrade order eligible.
SERVER GAMMA — STORAGE
94% utilization — threshold exceeded. Alert active. Duplicate records identified for local DR copy.
Threshold values are illustrative. The patent claims the detection and alerting mechanism; specific percentages are implementation choices.
08 / Practical applications
Where autonomous remediation earns its keep.
The patent expressly claims data center self-healing and local disaster recovery. The peer-remediation mesh and threshold monitoring suggest broader applications wherever infrastructure must respond autonomously to change.
Express applications are explicitly claimed. Inferred applications extend the mechanism to reasonable adjacent domains.
Application landscape
09 / Referenced by later patents
Cited by Microsoft, EMC — and 24 more.
Google Patents lists 26 forward citations. Microsoft cited the datacenter telemetry approach in 2024. EMC (Dell) cited it twice for IoT-driven preemptive datacenter measures. Digital Global Systems built a patent family of 22 continuations citing this work.
These are forward patent citations — a citation signals relevance to a later patent examiner or applicant, not necessarily commercial deployment.
Forward citations shown by Google Patents / checked 2026-06-10
Microsoft
EMC / Dell
Digital Global Systems
+ 23 more
26 total forward citations. Digital Global Systems accounts for 22 citations in a single patent family (spectrum management). Additional cites: State Grid Zhejiang (China). EMC is a subsidiary of Dell Technologies.
10 / Patent record
Filed and granted in thirteen months.
US10721135B1 defines edge computing as intelligence at the node — not in the cloud, not in a controller, but in the sensor permanently attached to each component. The durable insight: infrastructure that classifies its own changes can fix itself faster than any human on-call rotation.
Jun 2019
Priority Filed
US16/442,655 filed
priority date established
priority date established
1 yr 1 mo
Jul 2020
Granted
USPTO granted US10721135B1
16 claims, 2 independent
16 claims, 2 independent
~19 years
Jun 2039
Expires
Anticipated expiration
per Google Patents
per Google Patents
End / Patent 04