UPS Batteries: VRLA vs Li-ion, Replacement Cycles, Testing
How to compare VRLA and lithium ion UPS batteries, plan replacement cycles against temperature and load, and run the IEEE 1188 test cadence so batteries fail on your schedule instead of during an outage.

On this page
- At a Glance: VRLA vs Li-ion
- What Design Life Actually Means
- VRLA: Cheap to Buy, Expensive to Supervise
- Li-ion: Fewer Replacements, Different Responsibilities
- Counting Replacement Cycles Against the UPS Lifecycle
- Testing VRLA: The IEEE 1188 Cadence
- Testing Li-ion and Polling the Fleet from Your Desk
- The Annual Battery Audit
A UPS battery is the least reliable component in the power chain and the one most likely to be managed by memory. The UPS was sized carefully, the generator gets its NFPA 110 load tests, and the battery string hums on float in a cabinet until one day the runtime that was supposed to be ten minutes turns out to be ninety seconds. Battery failure is rarely sudden in the way people imagine it: a VRLA string often measures perfectly on float voltage while having quietly lost half its capacity, which means it passes every casual glance and then fails under the load of an actual outage. This article walks through the two chemistries you will actually be choosing between, how to plan replacement cycles against temperature and the UPS lifecycle, and the test cadence that turns battery health from a guess into a number.
The comparison matters because the two chemistries fail differently and cost differently. VRLA is cheap to buy and expensive to supervise, and it almost always needs at least one replacement inside a UPS lifecycle. Lithium ion costs more upfront, arrives with its own battery management system, and often needs no replacement at all. Which one is right depends on your room temperature, your floor space, how many refresh cycles you are willing to schedule, and how much test labor you want to commit to every year.
At a Glance: VRLA vs Li-ion
| VRLA | Li-ion (LFP) | |
|---|---|---|
| Typical service life | 3 to 5 years at 25°C | 8 to 10 years or more |
| Discharge cycles | 200 to 500 | 3,000 to 5,000 |
| Weight and footprint | Baseline | 40 to 60% lighter, roughly 40% smaller |
| Monitoring | Optional add-on (battery monitoring system) | Integrated battery management system per cell |
| Temperature tolerance | Loses about half its life per 8 to 10°C above 25°C | Far more tolerant, though life still drops above 30°C |
| Upfront cost | Baseline | 1.2 to 2x for the same energy |
| Test burden | Monthly through annual per IEEE 1188 | Mostly BMS self reporting, periodic verification |
| Replacements in a 10 year UPS | Usually 2 (at 25°C) | Usually 0 |
What Design Life Actually Means
The first distinction to get straight is calendar life versus service life. Calendar life is how long a battery lasts sitting on float with no outages at a stated temperature, usually 25°C. Service life is how long it lasts in real conditions, and that is the number that matters, because every discharge event, every hot afternoon in a marginal room, and every float voltage drift spends part of the budget. End of life for both chemistries is defined as the point where the battery can still deliver only about 80 percent of its rated capacity, not the day the UPS starts alarming. A battery that passed last year’s check at 85 percent of capacity is already in its last act even if the front panel is green.
Temperature is the single biggest lever you control, and the rule of thumb is Arrhenius: for every 8 to 10°C of average ambient above 25°C, remaining battery life roughly halves, a relationship that holds for VRLA and, to a lesser degree, for lithium ion. The effect is easy to misjudge because rooms drift. A network closet with a 1000 VA unit that was specified at 25°C and actually lives at 30°C is not losing 20 percent of its battery life, it is losing half of it.
| Average ambient | Life remaining vs rating | Why |
|---|---|---|
| 25°C | 100% | Manufacturer reference condition |
| 30°C | About half | Chemical reaction rate roughly doubles |
| 35°C | About a quarter | Second halving inside the same decade |
| 40°C | About an eighth | Also a safety margin problem for the enclosure |
This table is the reason the first line item in any battery program is a real temperature reading at the cabinet, not the thermostat in the hallway. Measure the average over a week, then recompute every life expectancy and replacement budget you have ever written down, because most of them were written against 25°C.
VRLA: Cheap to Buy, Expensive to Supervise
VRLA, valve regulated lead acid, is the sealed lead acid battery in most UPS units on the market, and its economics are the economics of the replacement cycle: the blocks themselves are inexpensive, but a string of them fails unpredictably, so someone has to watch it. Two VRLA failure modes dominate. The first is dry out: during float, oxygen recombines at the plates and consumes water inside the sealed cell, so the battery slowly loses the electrolyte it needs to hold charge. The second is thermal runaway: heat increases float current, float current generates more heat, and a block in a warm cabinet can run away until it swells, vents, or worse, which is why the room temperature matters as much as the charger setting.
The defense for both is disciplined float voltage. Most manufacturers specify 2.25 to 2.30 volts per cell at 25°C, which is 13.5 to 13.8 volts for a 12 volt block, ideally temperature compensated so the charger backs off as the room warms. The total string voltage can hide a sick block, so the quarterly check measures each block individually. A string can also sit at correct float voltage while having lost half its capacity, which is exactly why voltage alone is not a health test and why the ohmic testing cadence in the next sections exists.
The practical consequence is that VRLA needs a supervision program: quarterly ohmic readings, an annual connection resistance check, and a capacity test every two years or less. That labor is real, and it is the cost column that a lithium ion comparison often fails to itemize. VRLA remains the right call for short ride through requirements, tight budgets, and sites where the batteries are easy to reach and cheap to swap, provided someone actually owns the test schedule.
Li-ion: Fewer Replacements, Different Responsibilities
Lithium ion UPS batteries, most commonly lithium iron phosphate (LFP) chemistry in this application, change the economics by outliving the UPS. Service life runs 8 to 10 years or more, cycle life runs 3,000 to 5,000 discharges against 200 to 500 for VRLA, and the pack is 40 to 60 percent lighter and roughly 40 percent smaller for the same energy. The recharge is several times faster too, which matters after back to back outages: a VRLA string can take many hours to recover its charge while a lithium pack is ready again in a fraction of the time.
The structural difference is the battery management system. VRLA monitoring is an optional add-on that the data center operator chooses to buy; a lithium pack cannot ship without a BMS because the chemistry requires per cell control of charge voltage, current, and temperature, and the BMS disconnects the battery on unsafe conditions. That means state of charge and state of health are measured continuously rather than sampled quarterly, and the degradation data you need for a replacement decision arrives without a site visit. The tradeoffs are real: capex runs 1.2 to 2 times higher for the same energy, transport involves stricter regulations for the lithium cells, and while lithium tolerates heat far better than VRLA, published vendor guidance still warns that a 10 to 15 year calendar life shortens in rooms sustained above 30°C.
Over a 10 year horizon the math usually favors lithium. A widely cited Schneider Electric white paper (White Paper 229) modeled a 1 MW UPS at 25°C and found lithium’s total cost of ownership 39 percent lower than VRLA, with the savings driven almost entirely by the two VRLA battery refreshes and their associated maintenance, floor space, and cooling that lithium never incurs. Lithium pays back fastest in hot rooms, tight spaces, edge sites, and anywhere a battery swap means scheduling a maintenance window with an affected customer.
Counting Replacement Cycles Against the UPS Lifecycle
A replacement plan should be written against the UPS lifecycle, not the battery datasheet. The typical UPS serves 10 years, and a VRLA string specified at 4 years of service life needs a refresh around year 4 and again around year 8, which means two procurements, two installation windows, and two moments where a wrong torque or a bad block from a batch can walk into your plant. The same UPS on lithium expects no refresh at all. In a 30°C room the VRLA track doubles to roughly four refreshes, which is the scenario where the lithium premium pays back in under three years.
| Scenario | Expected battery life | Refreshes in a 10 year UPS | Budget shape |
|---|---|---|---|
| VRLA, 25°C room | 3 to 5 years | 2 | Two refresh procurements plus annual testing |
| VRLA, 30°C room | 1.5 to 2.5 years | About 4 | Refresh every other year, testing annually |
| VRLA, 35°C room | About 1 year | 8 or more | The room is the problem, fix cooling first |
| Li-ion, any of the above | 8 to 10 years or more | 0 | One procurement, BMS reports the rest |
Two planning rules keep this honest. First, never let a VRLA string reach its refresh year untested: IEEE 1188 requires a capacity test at least every two years for healthy strings and annually once the battery passes 85 percent of expected life, so by the time the refresh budget lands the test data should already have confirmed the decision. Second, budget runtime against an aging bank: a string at end of life delivers about 80 percent of rated capacity, so a UPS sized for ten minutes of ride through delivers closer to eight when you actually need it.
Testing VRLA: The IEEE 1188 Cadence
The hierarchy of UPS battery tests matters, because each level buys a different amount of certainty. A UPS self test is a short discharge that proves the battery is present and roughly alive; it will not find a string that has quietly dropped to 80 percent capacity. Runtime remaining, the value the UPS front panel shows, is load dependent and assumes a healthy battery, so it degrades with the battery without anyone noticing. Per block ohmic testing with a handheld impedance or resistance tester is the quarterly workhorse: it catches dried out cells, plate corrosion, and bad intercell connections by comparing each block against the baseline recorded when the string was new. Ohmic values are trending tools, not verdicts; the industry consensus in IEEE 450, 1188, and 1106 is that a controlled discharge test is the only measurement that truly establishes capacity, which is why IEEE 1188 requires it on a two year cycle for healthy strings, annually past 85 percent of expected life, and immediately when a string shows degradation, defined as a capacity drop of more than 10 percent from the previous test or a result below 90 percent of rating.
The replacement threshold is 80 percent of rated capacity. When a capacity test returns below 80 percent, the battery should be replaced, and IEEE 1188 allows at most one year to do it, because a string below that line no longer delivers the runtime the UPS was sized for. Treat the tripwire accordingly: a string that fails a capacity test at year 3 is not a “monitor it” finding, it is a purchase order with a deadline.
Testing Li-ion and Polling the Fleet from Your Desk
A lithium UPS needs almost none of the field test cadence above, because the BMS measures every cell continuously and reports state of health, temperature, and charge status to the UPS management interface. Your verification job shifts from quarterly site visits to reading the data the battery already produces: check the BMS reported state of health after each firmware update and each significant outage, confirm the UPS logs show no BMS protection events, and reconcile the runtime the UPS projects against the load you are actually carrying. What you cannot do is skip the environmental half of the program: the BMS protects the cells, but it does not fix a room that stays above 30°C and quietly shortens the pack’s calendar life.
For VRLA fleets, and for any UPS with a network management card, SNMP polling gives you a baseline battery audit that costs nothing but a few lines of shell. On APC hardware running the PowerNet MIB, the battery OIDs report capacity, temperature, runtime remaining, and a replace indicator directly:
snmpget -v2c -c "${COMMUNITY}" -Oqv nyc1-ups-a \
1.3.6.1.4.1.318.1.1.1.2.2.1.0 \
1.3.6.1.4.1.318.1.1.1.2.2.3.0 \
1.3.6.1.4.1.318.1.1.1.2.2.4.0 \
1.3.6.1.4.1.318.1.1.1.2.1.3.0
98
1740
1
03/14/25
Read the output in order: 98 is remaining capacity as a percentage, 1740 is runtime remaining in hundredths of a second (TimeTicks, so 17.4 minutes at the current load), 1 means noBatteryNeedsReplacing with 2 meaning the UPS wants the string swapped, and the date is the last recorded battery replacement, which the installer should set whenever blocks are changed. The replace indicator is set by the UPS’s own internal self test logic, so treat it as a second opinion that should agree with your test data, not as a replacement for it.
Hosts using the Network UPS Tools client expose the same information through upsc:
upsc nyc1-ups-a@monitor ups.status # OL on line, OB on battery, LB low battery
upsc nyc1-ups-a@monitor battery.charge # percent
upsc nyc1-ups-a@monitor battery.runtime # seconds
upsc nyc1-ups-a@monitor battery.voltage # DC volts
For a fleet, put each hostname in a file and run the same four OIDs into a dated CSV so every quarter is one diff against the last:
AUDIT="ups-battery-audit/$(date +%F)"
mkdir -p "${AUDIT}"
printf 'host,capacity_pct,temp_c,replace,last_replaced\n' > "${AUDIT}/fleet.csv"
while read -r host; do
snmpget -v2c -c "${COMMUNITY}" -Oqv "${host}" \
1.3.6.1.4.1.318.1.1.1.2.2.1.0 \
1.3.6.1.4.1.318.1.1.1.2.2.2.0 \
1.3.6.1.4.1.318.1.1.1.2.2.4.0 \
1.3.6.1.4.1.318.1.1.1.2.1.3.0 \
| paste -sd, | sed "s/^/${host},/" \
>> "${AUDIT}/fleet.csv"
done < ups-hosts.txt
awk -F, 'NR > 1 && ($2 < 90 || $4 == 2)' "${AUDIT}/fleet.csv"
gru1-ups-a,77,31,2,06/02/23
The one row that comes back is the whole point: capacity at 77 percent is below the 80 percent IEEE 1188 replacement floor, the UPS’s own replace indicator agrees, and the install date of June 2023 shows a string aging out early in a room that is averaging 31°C, which the temperature halving rule predicts. Keep one directory per run date, and diffing consecutive quarters shows you which strings are declining fastest, which is exactly the signal a monthly glance at a green front panel never gives you.
The Annual Battery Audit
Run the audit as a procedure, in this order, because each step produces the fact the next one needs:
- Generate the dated fleet CSV with the SNMP loop above, then diff it against the previous run. Every row that changed is your inspection list before you leave your desk.
- Walk each flagged UPS and record the real ambient temperature at the battery cabinet with a handheld meter, then compare it to the polled internal value. Recompute expected remaining life against the halving rule if the room is above 25°C.
- For VRLA strings, measure individual block float voltages with a multimeter. Most manufacturers want 13.5 to 13.8 volts per 12 volt block; blocks sitting outside the manufacturer’s spread on a float system are candidates for the ohmic tester.
- Review the quarterly ohmic trend. Any block deviating more than about 20 percent from its baseline, or a string whose readings shifted suddenly between quarters, triggers a capacity test rather than another quarter of watching.
- Reconcile the
last_replaceddates from SNMP against purchase records and your documentation. An install date that nobody set after a swap makes every aging calculation wrong, and fixing it costs one SET, not one site visit. - Schedule the capacity tests that fall due this year: acceptance for new installs, the two year performance test for healthy strings, and annual testing for anything past 85 percent of expected life or showing degradation.
- Close by re-planning the replacement budget against measured temperatures and the UPS lifecycle table, so the year’s refresh procurement is a line item and not a surprise.
| Finding | Evidence | Fix | Owner |
|---|---|---|---|
| Capacity 77%, replace indicator 2 | Fleet CSV row | Order the replacement string; below 80% must be replaced within a year | Facilities lead |
| Cabinet averages 31°C | Polled temperature plus handheld reading | Fix cooling or ventilation, then recompute life at half per 8 to 10°C | Facilities lead |
| Two blocks deviate 20% from ohmic baseline | Quarterly ohmic trend | Capacity test the string, replace weak blocks | Battery vendor |
| Runtime fell from 17 to 9 minutes at the same load | Runtime OID trend against load history | Treat the 80% floor as breached, test and replace | Facilities lead |
| Install date predates the actual swap | last_replaced vs purchase records |
Correct the record and the aging calculation | Documentation owner |
The lesson that runs through all of it: batteries fail on a schedule that is predictable only if you measure the inputs, temperature and capacity, and invisible if you rely on the front panel. Run the SNMP loop on one UPS today and look at the four numbers it returns; if the capacity percentage and the install date disagree with what you assumed, you have already found something.
For teams that already keep infrastructure documentation, recording each UPS as a power source with the battery install date in a custom field keeps the replacement clock next to the metered load the battery actually serves, so the refresh decision can be checked against data instead of memory. The data center management use case covers how that record fits together with panels, breakers, and feeds.
Frequently Asked Questions
How long do UPS batteries last?
How often should UPS batteries be replaced?
How do you test a UPS battery without a load bank?
Is lithium ion worth it for a UPS?
What float voltage should VRLA UPS batteries have?
How do I track battery install dates and test results for a UPS fleet?
Stop reaching for a spreadsheet
Obelinf keeps every subnet, device, circuit, and rack in one live source of truth, with audit logs and a topology view. Free for personal use.
Related Articles

How to Document Power Sources, Panels, Breakers, and Feeds
A working procedure for building a power distribution record from source to rack: the order to create each record, the API calls, the fields that matter, and the checks that catch gaps.
Read more
How to Prove A/B Power Redundancy in Practice
Use rack, feed, panel, and device power records to prove A/B redundancy: the commands, the failure domain counts, the common false positives, and a worked example.
Read more
What Is a Power Feed? Panels, Breakers, and Rack Power Explained
A plain explanation of power feeds, panels, breakers, circuits, and PDUs, with the fields to record, the capacity math, and a checklist you can use at the rack.
Read more