Plan Network Capacity with Confidence at Scale

Kentik’s Capacity Planning workflow automates bandwidth forecasting by analyzing SNMP interface metrics and historical traffic to compute both utilization and runout for any group of links. Instead of spreadsheet gymnastics, you get daily forecasts and automatic runout projections that show when specific links are likely to saturate. From there you can prioritize upgrades based on which interfaces will run out of capacity first and how critical they are to your business.

Right-sizing interconnect capacity requires trend analysis of measured peaks — not averages — per transit provider and IX port, projected forward against growth, with headroom policies that reflect the cost of congestion for that specific link (a saturated IX port silently degrades peering traffic; a saturated transit link costs money in overage or burst billing). The common failure modes are provisioning to averages and treating all links with the same headroom rule. Kentik supports this by tracking utilization and 95th-percentile peaks per interconnect, forecasting runout dates from measured growth, and attributing traffic on each link to its drivers — so upgrades are sized and prioritized by evidence, and expensive capacity isn’t added where traffic could be shifted instead.

Effective capacity planning requires a combination of interface-level metrics (utilization, error rates, discards from SNMP and streaming telemetry), flow-level metrics (traffic volume by application, source, destination, and AS path), and trend data (growth rates, seasonality, peak-to-average ratios). Kentik unifies these data sources in one platform — correlating SNMP interface metrics with NetFlow/sFlow/IPFIX flow data and BGP routing context — so capacity decisions can be made on the full picture rather than on partial information. For deeper context on the underlying telemetry, see Network Performance Monitoring on Kentipedia.

Early saturation detection means alerting on leading indicators rather than utilization ceilings: sustained growth in peak utilization trend, shrinking headroom windows during daily peaks, rising queue drops or errors on the interface, and shifts in traffic mix that predict faster growth. By the time utilization alarms at 95%, users are already experiencing the congestion. Kentik supports this by baselining utilization patterns per interconnect, forecasting when each link will hit critical thresholds based on measured growth, and alerting on trend — so capacity decisions happen weeks ahead of congestion, with traffic attribution showing whether the fix is an upgrade, a peering change, or traffic engineering.

Hybrid cloud capacity forecasting requires visibility into traffic across data centers, cloud regions, and the connections between them — not just on individual links. Kentik ingests flow data from on-premises infrastructure (NetFlow, sFlow, IPFIX) and from major cloud providers (AWS, Azure, Google Cloud, Oracle Cloud Infrastructure) into a unified platform, then applies the same forecasting and runout projections across every environment. This makes it possible to anticipate capacity needs on Direct Connect, ExpressRoute, Cloud Interconnect, and FastConnect circuits alongside on-premises WAN links, and to plan capacity in the context of how cloud and on-premises traffic interact.

Network baselining captures what “normal” looks like for each interface, application, and traffic flow over time — so deviations stand out clearly when they occur. Kentik automatically generates baselines from historical flow and metric data, then surfaces deviations that may indicate capacity issues (sudden bandwidth growth, unusual application traffic patterns, unexpected traffic from new sources) before they impact users. Combined with dynamic thresholds, this approach reduces alert fatigue compared to static percent-full thresholds and catches anomalies that would otherwise go unnoticed.

Major-event capacity management (World Cup streaming surges, game releases, product launches) combines forecasting from comparable historical events, pre-event validation that critical links and interconnects have headroom, and real-time monitoring during the event with pre-agreed mitigation options — traffic engineering shifts, CDN coordination, temporary capacity — ready to execute. The teams that handle events well rehearse the traffic scenario beforehand rather than discovering their bottleneck live. Kentik supports this by providing historical traffic analysis of past events for forecasting, per-link and per-interconnect headroom visibility before the event, and real-time anomaly detection during it — so operators can see hotspots forming and act before subscribers feel them.

Service provider capacity planning involves orders of magnitude more interfaces, traffic, and customer-specific context than enterprise capacity planning — and the analytics platform has to keep up. Kentik scales to ingest flow data from large backbone networks, peering edges, and customer interconnects, then applies forecasting and runout projections across all of it. Multi-tenant analytics let provider teams plan capacity for specific customers, services, or geographies; peering and transit analytics inform interconnection strategy alongside capacity decisions. For more on Kentik’s service provider capabilities, see Kentik for Service Providers.