On-Change vs Periodic Telemetry for Device State?
An infrastructure team is concerned about the shared memory utilization of a device, and for this reason, they need to monitor the device state. Which solution limits impact on the device and provides the required data?
Community Insight
The test is whether you recognize that operational state data (up/down, adjacency status) changes infrequently, so an on-change subscription keeps device load low, while the trap is choosing a time-based periodic subscription because it feels more 'continuous'.
Cisco model-driven telemetry offers periodic and on-change YANG subscriptions, and this question asks which one lets an infrastructure team monitor a device's state while limiting impact on shared memory. The page establishes that an on-change subscription is the correct answer because it transmits only when the subscribed state value actually changes.
Option D (periodic subscription) is the most common wrong pick — two commentators argue that memory figures 'change very frequently' and that time-interval reporting suits utilization sampling, but the question explicitly targets device state, not a numeric gauge, and periodic pushes data for the entire subscription lifetime regardless of need.
Community Discussion (18 comments)
Comments & Corrections
No comments yet — spotted an error or have a note? Share it below.
Expert Analysis
Why the Answer Is Correct
An on-change subscription publishes the initial object state once and then sends updates only when a subscribed value changes, so a device whose state is largely static generates almost no telemetry writes or buffer consumption. Because the team's stated concern is shared memory utilization, minimizing the number of pushed updates directly reduces pressure on the device's telemetry buffers and CPU. Operational state leaves (interface up/down, OSPF or BGP neighbor status) are exactly the leaf types designed for on-change detection, so the required data still arrives in near real time. This matches Cisco's model-driven telemetry subscription model, where on-change is the low-impact mechanism for monitoring device state.Why the Other Options Are Wrong
Option D, a periodic subscription, streams the full subscribed dataset at every configured interval for the lifetime of the subscription, which continuously consumes shared memory and CPU even when nothing has changed — precisely the impact the team wants to limit. Option A, IPFIX, is an IP flow export protocol tied to flow records and sampling, not a YANG subscription for device state, so it neither scopes the data to state nor offers the low-impact delivery described. Option B, static telemetry, refers to polled or fixed-configuration collection approaches rather than an event-driven subscription, and polling repeatedly walks the datastore instead of reacting only to change. None of these three deliver change-triggered, state-specific reporting with the smallest device footprint.Community Comment Notes
Most commenters converge on C, with goku2020 stating plainly that "C is correct Change-on subscription have less impact" and bogd reducing the logic to "Less impact ==> on-change subscription." akbntc and namibdigger both point to classic on-change use cases such as an interface or OSPF adjacency going down, while namibdigger cautions that this monitors "the amount of memory used which might work with A (in SDA)." The dissenting voices, zzmejce and CKL_SG, argue for D on the grounds that memory is measured in bytes and changes frequently, or that periodic reporting shows utilization "from time to time" — but both readings mistake a numeric gauge for the state data the question actually names. magicrushb76 adds a useful nuance that the first on-change push is an uncontrollable full synchronization, after which only changed rows are sent.Official Reference
Exam Strategy
Key off the noun in the question: 'state' (up/down, adjacency) points to on-change, while a numeric gauge or rate points to periodic sampling. Whenever a question stresses limited device impact, the event-driven option is almost always the intended answer.
Frequently Asked Questions
Why is a periodic subscription wrong when memory keeps changing?
The question asks to monitor device state (up/down, adjacency), not a numeric gauge. Periodic pushes the full dataset every interval regardless of need, raising shared memory and CPU load.
Could IPFIX be used to monitor shared memory utilization?
No. IPFIX exports sampled IP flow records, not YANG-modeled device state, so it does not provide the targeted, low-impact state data described in the scenario.
Related Analysis
Practice All 300-420 Questions
Access 150 questions with complete answers and detailed explanations.
View Full 300-420 Practice Test →