Connectivity and backup power for a bank branch on an unreliable ADSL line
A cooperative bank's branch in a tier-3 town runs nine counter terminals, one branch server hosting the core banking client, a four-camera IP surveillance set monitored from the regional office, and fourteen staff. Its only link is a 4 Mbps ADSL connection that drops for three to six hours during the monsoon. The regional office now requires six simultaneous video calls to head office at 2 Mbps each in both directions, a 1.2 GB nightly backup uploaded between 22:00 and 06:00, and counters that keep working for at least four hours during a power cut. Two connectivity quotes are on the table. Quote A: 100 Mbps FTTH, Rs 9,000 per month, advertised contention ratio 1:8, 99% uptime, 8-hour repair. Quote B: 20 Mbps 1:1 leased line, Rs 28,000 per month, 99.5% uptime, 4-hour repair. The branch can fund one primary link plus one cheap standby. Backup power is a 1 kVA online UPS (power factor 0.8) with two 12 V, 100 Ah batteries; measured critical load is 450 W.
1. The branch manager prefers Quote A because '100 Mbps is five times 20 Mbps'. Evaluate that comparison and recommend the primary and standby links.
Read the worked answer
Compare guaranteed rates, not advertised peaks. A 1:8 contention ratio on 100 Mbps FTTH means the committed rate at the busy hour is about 100 / 8 = 12.5 Mbps. The six video calls alone need 12 Mbps in each direction, so the FTTH link has roughly 0.5 Mbps of headroom for core banking postings, the camera feed to the regional office, the nightly backup and ordinary browsing. That is not a design, it is a coincidence. It is worse than it looks because GPON upstream is usually the weaker direction.
Quote B gives all 20 Mbps, symmetric and uncontended, leaving about 8 Mbps of headroom. Recommend the leased line as primary with FTTH as automatic standby: Rs 37,000 in total, and the FTTH 8-hour repair SLA is acceptable precisely because it is the standby, not the primary. The '100 is more than 20' argument compares an advertised peak against a guaranteed floor.
Fallback if the branch truly cannot fund both: FTTH primary plus a 4G router standby, with QoS reserving 12 Mbps for video and prioritising core banking above the backup job above general browsing. Cheaper, but video will stutter at peak and the branch should say so in writing.
2. Does the existing UPS meet the four-hour requirement? Show the calculation and give a concrete remedy.
Read the worked answer
UPS capacity: 1 kVA x 0.8 power factor = 800 W, so the 450 W load is at 56% loading, which is healthy (keep it under 80%).
Battery energy: 2 x 12 V x 100 Ah = 2,400 Wh. Usable energy = 2,400 x 0.8 depth of discharge x 0.85 inverter/DC-DC efficiency = 1,632 Wh. Runtime = 1,632 / 450 = 3.63 h, roughly 3 h 35 min. The requirement is 4 h, so it fails by about 25 minutes even with brand-new batteries.
To deliver 4 h you need 450 x 4 = 1,800 Wh usable, which requires 1,800 / (0.8 x 0.85) = 2,647 Wh of battery, i.e. roughly 110 Ah per 12 V block, and that is before derating for a battery at end of life (about 80% of rated capacity) and for the fact that a branch room runs above the 25 C at which capacity is quoted.
Remedy: add a second matched string of two 12 V 100 Ah batteries wired in parallel with the first (never add a single battery in series across a 24 V bank). That gives 4,800 Wh raw, about 7.2 h fresh and about 5.8 h at end of life. Alternatively replace the bank with a new 2 x 12 V 150 Ah set for about 5.4 h, or take the cameras off the UPS.
3. State a go-live acceptance test the branch should run before signing the contract, with the specific checks and pass criteria.
Read the worked answer
Test under the real peak load, not at 22:00 on a Sunday.
Load test, 10:00-12:00: six simultaneous video calls to head office, a continuous core banking posting loop, both camera streams and a backup job. Pass criteria: one-way latency under 150 ms, jitter under 30 ms, packet loss under 1%, no dropped calls. An idle speed-test screenshot proves nothing.
Failover test: pull the primary link during a live core banking session. Measure the time to failover, confirm the session survives or reconnects cleanly, and confirm staff know the manual dial-back procedure if it does not. Failover that works but takes four minutes is an outage.
Backup test: confirm the 1.2 GB job completes inside the 22:00-06:00 window on both links, then actually restore it to a test machine. At 20 Mbps a 1.2 GB job takes about 8 minutes; on a 512 kbps ADSL upstream it would take over five hours, which is why the old link was fragile.
Power test: a controlled discharge of the UPS at the measured 450 W load, logging runtime, plus a monthly battery health check.
Add monitoring for link status, UPS on-battery events and backup success, alerting the regional office rather than only the branch.