A data center operator discovers their 100 kVA UPS is only supporting 65 kW of actual server load because the UPS was sized in kVA while the IT equipment was specified in kW — and nobody accounted for the 0.90 power factor at the UPS input. UPS sizing errors cascade: undersized batteries provide inadequate runtime, undersized inverters overload during transfers, and undersized bypass circuits can't handle maintenance loads. Proper UPS sizing requires understanding the total critical load (in both kW and kVA), the required autonomy time, redundancy configuration, and future growth.

UPS topologies offer different protection levels. Offline (standby) UPS switches to battery during outages (4-12 ms transfer time) — adequate for desktop computers and non-critical loads. Line-interactive UPS adds voltage regulation via an autotransformer but still transfers to battery during outages — suitable for network equipment and small server rooms. Online (double-conversion) UPS continuously regenerates power through rectifier and inverter stages, providing complete isolation from all utility anomalies (zero transfer time, voltage regulation, frequency regulation, harmonic filtering). Data centers, hospitals, and financial institutions require online double-conversion UPS.

Battery runtime depends on battery capacity (Ah), discharge rate, end voltage, and temperature. Lead-acid batteries (VRLA/AGM) typically provide 5-30 minutes of runtime at full load, with useful life of 3-5 years for standard and 8-10 years for 'long-life' designs. Lithium-ion batteries offer 2-3× longer calendar life (10-15 years), 2-3× more charge cycles (3,000+ vs 300-500), higher energy density (smaller footprint), and faster recharge (2-4 hours vs 8-12 hours). Battery capacity decreases at higher discharge rates — a battery rated for 30 minutes at 50% load may only provide 8 minutes at 100% load (Peukert's effect).

Redundancy configurations determine UPS system availability. N configuration (single UPS, no redundancy) provides ~99% availability. N+1 (one more module than needed) provides ~99.99% by allowing one module failure without load impact. 2N (fully duplicated pathways) provides ~99.9999% — each path handles 100% of the load independently. Distributed redundancy configurations (e.g., 4 UPS each carrying 25% load) combine high availability with efficient loading. Tier rating per Uptime Institute: Tier I (N, 99.671%), Tier II (N+1, 99.741%), Tier III (2N with concurrent maintenance, 99.982%), Tier IV (fault-tolerant 2N, 99.995%).

UPS efficiency directly impacts operating costs and cooling requirements. A 500 kVA UPS at 93% efficiency wastes 35 kW as heat — requiring approximately 10 tons of additional air conditioning. Modern high-efficiency UPS designs achieve 96-97% in double-conversion mode and 99%+ in eco-mode (bypass mode with fast transfer). Eco-mode saves energy but reduces protection quality — the 2-4 ms transfer time during a utility event may cause IT equipment disruption. Some manufacturers offer 'eco-conversion' modes that provide near-double-conversion protection at 98-99% efficiency.

UPS maintenance and testing is governed by NFPA 111 and IEEE 450/1188. Monthly tests include visual inspection, battery voltage verification, and load transfer testing. Semi-annual: full-load discharge test to verify battery runtime. Annual: infrared thermography of connections, battery impedance testing, and full system functional test. Battery impedance testing can predict failures before they occur — a cell showing >20% impedance increase from baseline should be replaced. Most UPS failures in practice are battery failures — predictive maintenance prevents the most common cause of UPS-related downtime.