Benchmarks
What ruststream-zeromq costs on every message: throughput against the raw zeromq client, and
instructions and allocations of its code.
Against a raw client
The first chart runs each scenario three ways over real sockets. Raw client drives zeromq
sockets directly. Broker crate drives this crate's subscription with a loop of its own and no
runtime. RustStream is the service a user writes, a #[subscriber] handler in an app. The first
gap is what this crate costs over the client it wraps, and the second is what the runtime adds on
top.
Each loop is a bar from its best round to its slowest, with a dot at the best round and a notch at the median. The zero of the axis is the raw client's best round, and a dot right of zero is slower. A hollow dot is a loop whose difference from the raw client is smaller than the spread of the rounds. The table under the chart prints the best round with the median in parentheses.
Both scenarios are PUSH/PULL, over the two transports this crate runs on: tcp:// on the loopback
and ipc://. They differ in the path a message takes through the kernel and in nothing else, so
the pair shows how much of the cost is the socket.
The other end of every loop is a socket in the same process, because ZeroMQ peers talk directly, as Methodology explains.
Cost of the code
The second chart counts what one message costs on the service's thread: instructions under
callgrind and allocations under DHAT, in the steady state. Each scenario is a service on the
production broker, bound to a port on the loopback, with a peer on a thread of its own. The count
includes the framework's dispatch, this crate's code and the zeromq client's framing on that
thread. The peer's thread and the kernel are not counted. Cold start is what starting the
service, accepting the peer's connection and taking the first message cost once.
Results
- Raw client
- Broker crate
- RustStream
- rounds, best to slowest
- best round
- median round
- spread of the rounds
- indistinguishable
ruststream-zeromq 0.7.0 on ruststream 0.7.0, AMD Ryzen 9 7900X 12-Core Processor, Linux 7.2.10-x64v3-xanmod1-1, none: ZeroMQ has no broker, so the peer is a PUSH socket in the same process, over tcp:// on the loopback and over ipc://, with default socket options (the zeromq crate exposes no high-water mark), measured 2026-10-18. How the numbers were taken
- indistinguishable
- indistinguishable
- indistinguishable
- +3.4%
Show the numbers as a table
| Scenario | Raw client | Broker crate | RustStream | Overhead |
|---|---|---|---|---|
| PUSH/PULL over tcp:// on the loopback, 512 B JSON | 241,721 msg/s (237,164) | 238,327 msg/s (237,795) (indistinguishable over the client) | 241,545 msg/s (231,957) | indistinguishable |
| PUSH/PULL over ipc://, 512 B JSON | 972,020 msg/s (955,055) | 952,265 msg/s (934,673) (indistinguishable over the client) | 939,038 msg/s (936,838) | +3.4% |
AMD Ryzen 9 7900X 12-Core Processor, x86_64, 427-5912 MHz, 12 physical, 24 logical, 62.4 GiB, unknown, Linux 7.2.10-x64v3-xanmod1-1, 1.99.0, 3.25.1, bench, inheriting release (opt-level = 3, lto = fat, codegen-units = 16, debug = line-tables-only), ruststream-zeromq default (none), ruststream macros,json. ruststream-zeromq 0.7.0, ruststream 0.7.0, measured 2026-10-18.
Show the numbers as a table
| Scenario | Instructions | Allocations | Cold start |
|---|---|---|---|
| PUSH/PULL on the loopback, JSON decode into a small struct, ack each | 5,036.1 | 5 | 1,075,776 / 614 |
| DEALER/ROUTER on the loopback, answered to the requesting peer | 13,686 | 18 | 1,090,494 / 632 |
| PUSH/PULL on the loopback, in batches of 64 assembled on the client | 5,017.1 | 5 | 1,082,736 / 618 |
The machine
| Processor | AMD Ryzen 9 7900X 12-Core Processor |
|---|---|
| Architecture | x86_64 |
| Frequency | 427-5912 MHz |
| Cores | 12 physical, 24 logical |
| Memory | 62.4 GiB |
| Memory speed | unknown |
| Operating system | Linux 7.2.10-x64v3-xanmod1-1 |
| Transport | none: ZeroMQ has no broker, so the peer is a PUSH socket in the same process, over tcp:// on the loopback and over ipc://, with default socket options (the zeromq crate exposes no high-water mark) |
| Round trip | tcp:// 15.8 us, ipc:// 11.1 us (20000 REQ/REP exchanges on the same transport, outside every loop) |
| Compiler | 1.99.0 |
| Build profile | bench, inheriting release (opt-level = 3, lto = fat, codegen-units = 16, debug = line-tables-only) |
| Features | ruststream-zeromq default (none), ruststream macros,json |
| RUSTFLAGS | none (the recipe clears RUSTFLAGS, so the numbers are not tied to this CPU) |
| Valgrind | 3.25.1 |
The numbers come from docs/benchmarks/results.json of this version, written by the benchmark
recipe. Methodology says how they were taken, and Numbers what
they allow you to conclude.