Switching Capacity Calculator
Estimate full-duplex fabric bandwidth, packet forwarding rate, oversubscription, uplink aggregation, line-rate PPS, and buffer burst headroom for mixed-speed home lab switches.
Switch presets
Port mix and traffic model
Switch capacity result
Capacity breakdown
Switch fabric spec grid
Typical non-blocking fabric with about 11.9 Mpps and shallow shared buffers.
Common smart access switch fabric, roughly 65 Mpps at 64-byte frames.
Compact multigig switches need enough PPS for Wi-Fi 6/7 small packets.
Useful storage aggregation shape when east-west traffic can run hot.
Line-rate PPS table
| Port speed | 64-byte one-way PPS | Full-duplex PPS | Full-duplex fabric per port |
|---|---|---|---|
| 1 Gbps | 1.488 Mpps | 2.976 Mpps | 2 Gbps |
| 2.5 Gbps | 3.720 Mpps | 7.440 Mpps | 5 Gbps |
| 5 Gbps | 7.440 Mpps | 14.881 Mpps | 10 Gbps |
| 10 Gbps | 14.881 Mpps | 29.762 Mpps | 20 Gbps |
| 25 Gbps | 37.202 Mpps | 74.405 Mpps | 50 Gbps |
| 40 Gbps | 59.524 Mpps | 119.048 Mpps | 80 Gbps |
| 100 Gbps | 148.810 Mpps | 297.619 Mpps | 200 Gbps |
Minimum Ethernet frame rate includes 64-byte frame, 8-byte preamble/SFD, and 12-byte inter-frame gap on the wire.
Oversubscription guide
| Ratio | Access switch meaning | Storage fabric meaning | Practical note |
|---|---|---|---|
| 1:1 to 2:1 | Very strong | Preferred | Good for backups, NAS, and east-west VM traffic. |
| 3:1 to 4:1 | Normal home lab | Watch closely | Fine when most client ports are idle most of the time. |
| 5:1 to 8:1 | Busy edge | Risky | Use QoS and avoid many simultaneous high-rate flows. |
| Above 8:1 | Likely bottleneck | Avoid | Add uplinks, move heavy servers, or split the access layer. |
Buffer and uplink planning table
| Scenario | Suggested buffer | Uplink target | Why it matters |
|---|---|---|---|
| Basic 1G access | 1 to 4 MB | 1x to 2x10G | Short client bursts, light NAS access, and web traffic. |
| Wi-Fi 6/7 multigig | 8 to 16 MB | 2x10G+ | Many wireless clients can burst into a small uplink set. |
| 10G NAS cluster | 12 to 32 MB | 40G or 4x10G | Backups and VM storage create synchronized bursts. |
| 25G storage leaf | 32 MB+ | 100G or 4x25G | High throughput plus incast needs deeper queues. |
Common switch builds
| Build | Port mix | Fabric spec to seek | Forwarding spec to seek |
|---|---|---|---|
| Home office edge | 8x1G + 1x2.5G | 20 Gbps+ | 15 Mpps+ |
| PoE access layer | 24x1G + 2x10G | 88 Gbps+ | 65 Mpps+ |
| Homelab aggregation | 8x10G + 2x40G | 320 Gbps+ | 238 Mpps+ |
| Storage mini leaf | 8x25G + 2x100G | 800 Gbps+ | 595 Mpps+ |
Capacity tips
Fabric tip: Compare the vendor switching capacity against the full-duplex sum of every port you intend to light up. A 24x1G switch with two 10G uplinks is normally advertised near 88 Gbps because each port can send and receive at the same time.
Uplink tip: LACP improves aggregate throughput across many flows, but a single TCP flow usually lands on one physical member. Keep heavy NAS, backup, and virtualization flows close to the switch or give them faster uplinks.
It’s switch fabric, something you likely never considered when your backup job froze the third time this week. It’s fine: traffic light is green, the link status say everything is a go… yet throughput is low. Why? Too many cars on the freeway inside box. That’s switching capacity; the place where most home lab builders fall flat on their faces. You see numbers on ports and LEDs and ask yourself why your network feels like it runs through mud when you transfer lot of data.
Define your port mix in the calculator above and it’ll do the math for you. No need to guess if that shiny new box can absorbs what you’re asking it to do.
How to Check If Your Switch Is Fast Enough
There’s a difference between packet rate and bandwidth. The latter measure how much data passes through in a given period of time; the former measures how many individual envelopes the switch can open and sort in a second. More often than not, small packets will kill a forwarding engine sooner then eat up bandwidth. If your network runs mostly web browsing or VoIP, those tiny frame add up quickly. Those tiny frames add up fast. Sure, the switch may have tons of raw gigabits of throughput, but it doesn’t necessarily has enough cycles to handle sheer number of micro-transactions.
That’s where average frame size comes in. The tool converts what kind of traffic you expect into packets per second, which lets you compare to the vendor’s advertised forwarding rate. If you’re demanding more than that Mpps rating, then frames never make it onto the wire, they drop first.
It is the actual internal switching fabric that connects all the ports together. This is what vendors talk about when they say something like “this switch has 40Gbps.” Typically it’s assumed that each port sends and receives at the same time, which means number advertised by vendor is really a full-duplex capacity for each port. Add them all up and that’s how many total gigabits per second of internal bandwidth you’ll get from your switch. Two ten-gigabit uplinks plus twenty-four one-gigabit ports = about eighty-eight gigabits of internal bandwidth. Less than that, and there isn’t enough internal bandwidth to wire-speed connect all of your ports at once. That doesn’t mean the switch is broken. Somebody has to wait, it’s just a matter of who waits and how often.
Reality bites. Design follows. In an ideal world, your downlink capacity perfectly matches your uplinks. But such a thing doesn’t come cheap. You take what you get since three to one or four to one is typically good enough for access ports. After all, not all of your clients max out their connections at the same time, right? That’s why that ratio work.
Storage traffic is another matter entirely. Virtualization hosts and NAS arrays are prone to shout in unison. Oversubscribe them too much here and you’ll notice the lag. The page has a reference table that clearly spells this out. It tells you how much headroom to give yourself depending on your type of workload.
Then there are Uplinks, which add more trouble. Link Aggregation Control Protocol isn’t magic. It hashes flows and spreads traffic over various physical cables, but it doesn’t magically turn a big, fat upload into two or more big, fat uploads. A single big transfer typically stays on one cable no matter how many others exist. When planning, you need to take that asymmetry into consideration. The calculator takes this efficiency factor in account. What you see in its results is what’ll actualy hit the upstream router.
Packets in, packets out; input must exceed output; buffers soak up brief moment between. When your burst exceeds your buffer depth (packets dissapears into thin air). Switches these days have good buffers, but they come at a cost (in silicon). Reserving fifteen percent of all buffer, fabric, and forwarding requirements will keep things breathing easy.
As it turns out, counting the number of ports on the front is about as important as knowing how much room there is left inside the box. You don’t want to constantly babysit things for them to work. Before buying, check your capacity metrics so you don’t watch your network choke under load. You should of checked earlier.



