Editorially independent. Sponsors are disclosed and never influence our analysis.
Independent research, supported bycriblSponsor
Reference / Conversion

EPS to GB conversion for SIEM pricing: 2026 tables and worked examples

Independent reference for converting between Events Per Second and gigabytes per day across SIEM vendor pricing models. Interactive EPS-to-GB calculator, per-source conversion tables, three worked examples, and the discipline that prevents wrong vendor decisions. Updated September 2026.

Mixed enterprise
~70-100 EPS / GB
Typical baseline
Network-heavy
~200-400 EPS / GB
Firewall, NetFlow, DNS
Cloud-native
~30-50 EPS / GB
CloudTrail, MS365 audit
Sample period
30-60 days
Before any vendor sizing
Quick answer

There is no fixed EPS-to-GB ratio: it depends entirely on your log source mix. A mixed-enterprise environment averages roughly 70-100 events per second per GB per day (midpoint about 85 EPS/GB); network-heavy mixes run 200-400 EPS/GB and verbose cloud-audit logs run 30-50 EPS/GB. Enter your own figures in the calculator below, or use the reference table for a first-pass estimate at the 85 EPS/GB midpoint.

1 GB/day
~85 EPS
10 GB/day
~850 EPS
25 GB/day
~2,125 EPS
50 GB/day
~4,250 EPS
100 GB/day
~8,500 EPS
250 GB/day
~21,250 EPS

Modelled at the 85 EPS/GB mixed-enterprise midpoint (GB per day multiplied by 85). Your real ratio varies 30-50 percent with source mix; sample 30-60 days before any vendor sizing.

EPS to GB calculator
Set your log-source mix, then convert events-per-second and GB-per-day in either direction.
100%

Percentages are each source's share of total GB ingested. Need not sum to 100; the ratio is normalised.

Sustained event rate
5,925EPS
Weighted ratio
~119 EPS/GB
GB / month
1,500 GB
Mix in use: Windows event (security) 60%, Firewall (perimeter) 20%, Microsoft 365 audit 10%, EDR telemetry 10%.
First-pass only. Per-source ratios are typical observed midpoints; real environments vary 30-50% either way. Sample 30-60 days of your own EPS and GB-per-day before any vendor sizing.

The conversion problem

SIEM vendors split into two pricing camps based on the underlying cost driver of their architecture. Per-GB-priced SIEMs (Splunk, Microsoft Sentinel, Datadog, Devo, CrowdStrike LogScale) meter on data volume because their analytics engines scale with bytes processed; Sumo Logic's Flex model is the exception, billing scan and storage rather than ingest. Per-EPS-priced SIEMs (IBM QRadar, Securonix, LogRhythm via MPS) meter on event rate because their correlation engines scale with events per second. Both models are honest reflections of what costs money on the vendor side, but they make cross-shop comparisons require a conversion step.

The conversion is not a constant. EPS measures the rate of events per second; GB measures the volume of bytes ingested. The two correlate loosely but not tightly because event size varies enormously by source. A Windows security event log entry runs 1-2 KB on average; an AWS CloudTrail JSON event runs 5-15 KB; a firewall log entry runs 200-500 bytes; a NetFlow record runs 100-300 bytes. The same data volume in gigabytes produces wildly different event counts depending on which sources dominate the mix.

The practical effect is that customers comparing vendors based on assumed conversion ratios routinely get wrong answers by 30-50 percent in either direction. A network-heavy environment shopping per-EPS QRadar against per-GB Splunk based on a 100 EPS-per-GB conversion (the rough mixed-enterprise average) will materially under-buy QRadar capacity and over-buy Splunk capacity. The reverse profile (cloud-native, JSON-heavy) will produce the inverse error. The conversion discipline matters because the vendor decision frequently turns on it.

Per-source EPS-to-GB conversion table

Log sourceEPS per GBNotes
Windows Event Log (security)60-80 EPS / GBIncludes login, audit, system events
Windows Event Log (verbose / debug)200-350 EPS / GBService Control Manager spam, debug
Linux syslog (production)70-120 EPS / GBStandard syslog facility output
Firewall (perimeter)200-400 EPS / GBPer-rule logging, allow + deny
NetFlow / sFlow / IPFIX300-600 EPS / GBPer-flow records, very high event rate
Cloud-trail (AWS CloudTrail)40-80 EPS / GBVerbose JSON, low EPS, high GB
Azure Activity Log40-70 EPS / GBSimilar verbose JSON pattern
Microsoft 365 audit30-50 EPS / GBDetailed audit records, low EPS
EDR telemetry (CrowdStrike, SentinelOne)100-150 EPS / GBProcess trees, file events
DNS query logs400-700 EPS / GBVery high event rate, small per-event
Web proxy / SWG150-250 EPS / GBPer-request logging
Database audit (SQL Server, Oracle)80-140 EPS / GBPer-transaction audit records

Typical observed ratios across enterprise environments. Real-world variance of 30-50 percent in either direction is common. Use as first-pass reference, not as substitute for actual sampling.

Three worked examples

Mid-market with mixed sources

Source mix: 60% Windows event (~70 EPS/GB), 20% firewall (~300 EPS/GB), 10% MS365 (~40 EPS/GB), 10% EDR (~125 EPS/GB)

Weighted ratio: ~120 EPS / GB total

A 50 GB-per-day environment ingests roughly 6,000 EPS sustained. Splunk at 50 GB per day plus Enterprise Security runs roughly $100K all-in (about $50K base ingest, which ES roughly doubles), and that figure does not move when the event rate does. QRadar cannot be put on the other side of that line: IBM publishes no per-EPS rate, and the only public QRadar price is an AWS Marketplace contract covering 500 EPS, twelve times smaller than this profile. What the conversion does tell you is the direction of travel. At 120 EPS per GB this mix is event-dense, so an event-rate meter bills you for the same 50 GB more heavily than a byte meter does. Get IBM to quote 6,000 EPS and compare it against the $100K yourself.

Cloud-native engineering team

Source mix: 60% AWS CloudTrail (~60 EPS/GB), 15% MS365 (~40 EPS/GB), 15% EDR (~125 EPS/GB), 10% Linux syslog (~95 EPS/GB)

Weighted ratio: ~70 EPS / GB total

A 50 GB-per-day environment ingests roughly 3,500 EPS sustained. Splunk at 50 GB per day plus Enterprise Security runs roughly $100K all-in (about $50K base ingest, which ES roughly doubles). We will not put a QRadar number beside it, because IBM publishes none and the one public QRadar contract covers 500 EPS, a seventh of this profile. Structurally this is the mix that flatters an event-rate meter: at roughly 70 EPS per GB the events are large and comparatively few, so the event-rate axis is working in your favour rather than against it. That makes it the profile most worth getting an IBM quote for, and the choice then turns on content depth and source mix rather than a price win either way.

Network-heavy / regulated finance

Source mix: 50% firewall (~300 EPS/GB), 25% NetFlow (~450 EPS/GB), 15% Windows event (~70 EPS/GB), 10% DNS (~550 EPS/GB)

Weighted ratio: ~330 EPS / GB total

A 50 GB-per-day environment ingests roughly 16,500 EPS sustained. Splunk at 50 GB per day plus Enterprise Security runs roughly $100K all-in, and that figure is unchanged by the event rate. This is the profile where the two meters diverge hardest: at roughly 330 EPS per GB you are pushing thirty-three times the event count of the 500 EPS footprint IBM has publicly priced, on the same 50 GB. We cannot tell you what IBM will charge for that, and we will not guess, but a per-EPS meter is structurally the wrong shape for a network-heavy mix and this is the case to price carefully before committing.

The discipline that prevents wrong vendor decisions

Sample your environment for 30-60 days before any vendor sizing conversation. Modern SIEMs (any of the vendors listed in our pricing pages) provide native EPS and GB-per-day metering at the source level, free of charge during evaluation. Many environments already have this data available from an existing log management tool, even if the SIEM purchasing decision is brand-new. Without 30-60 days of source-level data, every vendor sizing conversation is based on assumption rather than measurement, and assumptions in this domain routinely produce wrong vendor decisions.

Calculate the weighted-average EPS-to-GB ratio across your real source mix. The math is straightforward once the sampling exists: for each major source, multiply its observed EPS by its observed GB-per-day, sum the products, divide by total GB-per-day. The resulting ratio applies to your specific environment and replaces the rough mixed-enterprise average that vendor pre-sales materials routinely use.

Apply the ratio in both directions when cross-shopping. Per-EPS vendor quotes get converted to per-GB equivalents using your environment's actual ratio; per-GB vendor quotes get converted to per-EPS equivalents using the inverse. The honest comparison shows where each vendor sits on cost in your specific environment, not in a generic environment that does not exist. Vendors that claim to be cheapest at the headline rate frequently are not cheapest in your specific source mix; vendors that look uncompetitive at the headline rate frequently win once the conversion is applied honestly.

Finally, never accept a vendor's own conversion ratio without sourcing it back to your actual environment data. Splunk pre-sales materials assume per-EPS conversion rates that overstate Splunk's competitive position; QRadar pre-sales materials assume rates that overstate QRadar's. The conflict of interest is real and unavoidable. The honest conversion is the one calculated from your data, not from any vendor's marketing.

FAQ

Common questions

Is there a calculator to convert EPS to GB per day?

Yes. The interactive EPS-to-GB calculator at the top of this page converts in both directions. Set the share of total GB ingested for each log source (or pick a mixed-enterprise, cloud-native, or network-heavy preset), enter either GB per day or sustained EPS, and it returns the converted figure plus the weighted EPS-per-GB ratio for your mix. The per-source ratios are typical observed midpoints, so treat the result as a first-pass estimate and confirm with 30-60 days of your own source-level sampling before committing to vendor sizing.

How many EPS is 1 GB per day?

There is no single fixed figure because event size varies by source, but a useful first-pass estimate uses the mixed-enterprise midpoint of about 85 events per second per GB per day. On that basis 1 GB per day is roughly 85 EPS sustained, 10 GB per day is roughly 850 EPS, 50 GB per day is roughly 4,250 EPS, and 100 GB per day is roughly 8,500 EPS. These are modelled at the 85 EPS/GB midpoint; a network-heavy mix (200-400 EPS/GB) produces two to four times as many events for the same GB, while a verbose cloud-audit mix (30-50 EPS/GB) produces fewer. Sample 30-60 days of your own source-level data before using any of these figures for vendor sizing.

How do I convert EPS to GB for SIEM vendor comparisons?

The conversion ratio depends entirely on your log source mix. Mixed enterprise log volumes average roughly 70-100 EPS per GB; network-heavy environments run 200-400 EPS per GB; verbose JSON cloud logs run 30-50 EPS per GB. To convert your environment, sample EPS and GB-per-day for each major log source over 30-60 days, calculate the weighted average ratio, and apply to compare per-EPS QRadar pricing against per-GB Splunk pricing. Assumed conversions routinely produce wrong vendor comparisons by 30-50 percent in either direction.

Why does the EPS-to-GB ratio vary so much by source?

EPS measures the rate of events per second; GB measures the volume of bytes ingested. The two correlate loosely but not tightly because event size varies enormously by source. A Windows security event log entry runs 1-2 KB on average. An AWS CloudTrail JSON event runs 5-15 KB. A firewall log entry runs 200-500 bytes. A NetFlow record runs 100-300 bytes. The same data volume in gigabytes produces wildly different event counts depending on which sources dominate the mix. Per-EPS pricing models (QRadar, Securonix, LogRhythm via MPS) reward small-event sources; per-GB pricing models (Splunk, Sentinel) reward large-event sources.

Which pricing model is cheaper for a network-heavy environment?

Per-GB pricing has the structural advantage for network-heavy environments where firewall, NetFlow, and DNS sources dominate the mix. The high-EPS, low-bytes-per-event characteristic of network logs punishes per-EPS billing: a 50 GB-per-day environment that is 75 percent network-source ingest produces roughly 15,000 EPS sustained, so an event-rate meter bills thirty times the event count of a 500 EPS footprint for the same 50 GB, while a per-GB vendor such as Splunk stays at roughly $100K all-in (50 GB per day plus Enterprise Security) regardless of event rate. We cannot give you the other half of that comparison in dollars, because IBM publishes no per-EPS rate for QRadar and the only public QRadar price covers 500 EPS. The conversion math still makes the point: run it, then get the event-rate vendor to quote your actual sustained EPS before you decide.

Which pricing model is cheaper for a cloud-native environment?

Per-EPS pricing has a structural edge for cloud-native environments dominated by verbose JSON audit logs (AWS CloudTrail, Azure Activity Log, Microsoft 365 audit), because the low-EPS, high-bytes-per-event characteristic of cloud audit logs is exactly what per-EPS billing rewards relative to a network-heavy mix. A 50 GB-per-day environment that is 70 percent cloud-audit-log ingest produces only about 3,500 EPS sustained, against roughly 15,000 EPS for the same volume of network logs, so the event-rate meter is counting far less for the same bytes. Whether that converts into a lower bill than Splunk at roughly $100K all-in (50 GB per day plus Enterprise Security) we cannot tell you, because IBM publishes no per-EPS rate for QRadar. The per-EPS advantage here is structural, relative to a network-heavy mix, and it is not the same thing as a headline price win over per-GB.

How accurate are the per-source EPS-to-GB ratios in your tables?

The ratios in our tables represent typical enterprise observations averaged across multiple environments and source configurations. Real-world ratios for any specific environment can vary by 30-50 percent in either direction depending on logging verbosity settings, event filtering at the source, message format compression, and similar implementation choices. The tables are useful for first-pass vendor comparisons and budget sanity checks; final vendor sizing should always be based on 30-60 day sampling of the actual environment, not assumed ratios.

Didn't find your answer?

Ask us. A real person reads every question and we answer the ones we can, with sources. If your question would help other readers, we may publish an anonymised version, with your permission. General reference only.

Updated 13 July 2026