About 50 workloads per load balancer
Each managed workload adds a target group, and a load balancer holds at most 100 (an AWS limit). The onboarding preview refuses when there is no room.
Scale
Tested at 1,000 accounts
What we measured on our test organization, how, the limits as they stand, and what Dormeto itself costs to run in your account.
- 67 sfor discovery to process a synthetic organization of 1,002 accounts and 17,033 account-regions
- 312 msmedian to page through 9,001 candidates, 200 at a time
- ~5 minfor a load balancer change to show, with event-driven refresh
What was tested
On our development environment, in October 2026. "Synthetic" data goes through the real pipeline and storage. "Simulated" runs the real code against fakes with AWS latencies and limits. "Live" is real AWS.
| Test | Result | How |
|---|---|---|
| Discovery of an organization | 1,002 accounts and 17,033 account-regions (16,030 without a load balancer) in 67 s, none failed. 10,000 candidates saved. A second run rewrote no unchanged candidate. | Synthetic |
| Candidates in the dashboard | 9,001 candidates in 46 pages, median 312 ms, at most 526 ms. | Synthetic |
| A crowded router | 1,500 workloads on 300 load balancers. Each idle check stays within its 240 s budget, and every workload is reached in 2 to 3 checks. Under heavy ELB throttling, it is slower but nothing fails. | Simulated |
| A burst of wakes | 200 wakes within 30 s against ELB's limit of 3 changes a second: every rule switch done, none failed. | Simulated |
| Bulk onboarding | Six onboardings on one load balancer at once: 3 at a time, all done in 57 s. Six offboardings in 39 s. | Live |
| Event-driven refresh | A new load balancer rule listed as a candidate within about 5 minutes, without a discovery run. Its removal likewise. | Live |
| Drift repair | A shadow rule deleted by hand put back in 69 s. | Live |
Limits as they stand
Lambda concurrency is shared
Held requests use your account's Lambda concurrency in the region. Each router's health warns when it may need more than 80% of the limit. New accounts start low: ask AWS for more.
ELB changes are shared
Rule switches share ELB's 3 changes a second with your own tools. Beyond about 500 wakes in 3 minutes in one account and region, a wake can fail and the next request retries it.
Naps 10 at a time per router
Thousands of workloads idle at once in one account and region nap over hours, not minutes. They stay awake meanwhile.
What Dormeto costs to run
Dormeto's own AWS charges in your accounts, not counting your load balancers nor your workloads.
- ~$0.33per load balancer and month: 1,000 load balancers in 100 accounts, 2 workloads each (estimate)
- ~$0.13per workload and month, its naps and wakes included
- ~$0.26per router and month, whatever its workloads
The cost follows routers and workloads, not load balancers. A workload is one target group of a load balancer, with every rule that forwards to it: what naps and wakes. A router naps and wakes the workloads of one account and region, however many load balancers they sit behind. A load balancer can only invoke a Lambda function of its own account and region. The manager adds about $5 a month. So a month costs about $5 + $0.26 per router + $0.13 per workload.
| 1,000 load balancers in 100 accounts | A month | Per load balancer |
|---|---|---|
| 1 workload each, 2 regions per account (200 routers) | ~$190 | ~$0.19 |
| 2 workloads each, 2 regions per account | ~$330 | ~$0.33 |
| 5 workloads each, 2 regions per account | ~$730 | ~$0.73 |
| 2 workloads each, 3 regions per account (300 routers) | ~$350 | ~$0.35 |
Estimates from AWS list prices (eu-west-1) and our measurements. They assume an idle check every 5 minutes, half the workloads awake and 3 nap and wake cycles per workload per working day. 10 cycles a day bring 2 workloads per load balancer to about $0.76. Your load balancers keep running and keep their own charges (about $18 a month each in eu-west-1): they receive the first request.