sysmond, the daemon that fills in Activity Monitor's numbers
It is the daemon behind Activity Monitor numbers: root, on demand, a minute and a quarter of CPU a day, zero log lines in six hours, the Disk tab column labels in its strings.
By Chethan Kumar S N · Updated
You opened Activity Monitor to find out what was making the Mac slow. It worked: ten columns of numbers about every process on the machine, one of them clearly the culprit. Now look at the list again, because one of the rows in it is the process producing the list.
That row is sysmond. It is the daemon that collects the per-process statistics the app you are staring at displays, the counter behind the counters, and this page is a measurement of it on a Mac running macOS 27.0.1 with 29 days of uptime. Everything here was read, nothing installed, nothing written.
One sentence of documentation, since 2010
Start with the manual page, because it is the shortest in this entire series:
SYSMOND(8) System Manager's Manual SYSMOND(8)
NAME
sysmond - system monitor daemon
DESCRIPTION
The sysmond command should not be run directly.
SEE ALSO
ps(1), top(1)
macOS 27.0 May 11, 2010 SYSMOND(8)
A name, a synopsis, one sentence of description, and a date from sixteen years ago. The see-also line is doing quiet work: ps and top, the two classic Unix tools that read process statistics. The daemon is named in the same breath as the tools it modernises.
Apple’s own Activity Monitor help book, fourteen articles strong, never returns the favour. This run fetched the book’s welcome page plus the four articles that matter, the ones on viewing process information, CPU activity, energy consumption and update frequency, and counted zero mentions of sysmond in any of them. The app describes its columns in detail and never names the process that gathers them. That gap is why the searches exist: what is sysmond on mac, what is sysmond in activity monitor, sysmond mac cpu, asked by people who found the name in the one app guaranteed to be showing it to them at the time.
The proof is in the load commands
The claim that this daemon feeds Activity Monitor is not a guess from the name. It is visible in the binaries. Activity Monitor’s executable lists /usr/lib/libsysmon.dylib among the libraries it loads. sysmond’s own executable lists the same library, plus two more that tell you where the Energy tab comes from, libpmenergy.dylib and libpmsample.dylib, and IOKit for the hardware counters underneath.
Then the control group, which is the interesting part. top, spindump, footprint and powermetrics, the command line tools that report the same kind of statistics, link none of these libraries. Zero. There are two architectures for building a monitoring tool on a Mac, sitting side by side in /usr/bin and /Applications: the old way, where the tool walks the kernel’s process tables itself, which is what top and its descendants do, and the new way, where the tool asks a daemon, once, through a Mach service named com.apple.sysmond, and receives tidy rows back. The command line kept the old way. The app Apple ships everyone uses the new way.
The vocabulary it ships
The strings inside the 273,568 byte binary are the daemon’s object model, and the model is a table:
OS_sysmond_object
OS_sysmond_process
OS_sysmond_coalition
OS_sysmond_system
OS_sysmond_request
OS_sysmond_row
A request comes in, rows go out, and each row is a process or the system as a whole. One class in that list explains a feature most people have used without knowing it had a name. Coalition is the mechanism that groups an app with all the helper processes it spawned, XPC services, renderers, extensions, so that the Energy tab can bill the bundle to the parent app instead of scattering it across the list. When Activity Monitor offers Applications in last 12 hours as a view, it is offering to collapse the process table into coalitions. The grouping you see is this daemon’s data structure.
The rest of the vocabulary is recognisable too. The strings include the literal column labels Bytes Read, Bytes Write, Operations Read and Operations Write, the Disk tab’s headers, the word Statistics, and the name of a source file this series has met before, process-libtop-memory.c, with its function libtop_update_vm_regions: the memory region walker from libtop, the library inside top, has been embedded in the daemon. The old tool’s counting engine was lifted into the new tool’s server. Alongside them sit the sysctls it reads to describe the machine itself, hw.memsize, kern.argmax, sysctl.proc_cputype.
The smallest launch file in the series
Ninety eight daemons into this series, and this is the leanest launch file measured yet, six keys and no schedule at all:
EnablePressuredExit => true
EnableTransactions => true
Label => com.apple.sysmond
MachServices => { com.apple.sysmond => true }
POSIXSpawnType => Adaptive
Program => /usr/libexec/sysmond
No RunAtLoad, no interval, no calendar, no gates. It runs as root from the system domain, a LaunchDaemon with a single door, and the door is the whole story. launchd’s live report for the job:
state = running
active count = 2
runs = 4
immediate reason = ipc (mach)
minimum runtime = 10
exit timeout = 5
last exit reason = JETSAM_REASON_MEMORY_IDLE_EXIT
jetsam memory limit (active, soft) = 30 MB
jetsam memory limit (inactive, soft) = 30 MB
jetsam priority = 140
jetsam thread limit = 32
Read the middle lines as a biography. Four runs since the Mac booted 29 days ago, each one ending the same deliberate way, the system taking the memory back while nobody was asking questions, each next question paying for a fresh start. The immediate reason line, ipc, means it is awake right now because a client is connected, two of them in fact, per the active count. A daemon whose entire existence is other people’s curiosity.
The five second question
The one place Apple documents this daemon’s behaviour is the update frequency setting, without naming it. The help article for it says: “By default, the information is updated every 5 seconds, but you can update information more frequently for more precise monitoring,” and then, unusually for a settings page, it warns you and tells you how to watch the cost:
“Increasing the update frequency may affect your overall system performance. To see how much processor time is being used to update Activity Monitor information, watch the CPU Time column in the CPU pane of the Activity Monitor window as you change the update frequency.”
That is a supported experiment, sanctioned in Apple’s own help, and it needs no tools. Open Activity Monitor, set Update Frequency to 1 second, and sort by CPU Time. Two processes accumulate time while you watch: Activity Monitor, for drawing the table, and sysmond, for filling it. Set the frequency back to 5 and both slow down. The monitoring has a price, the price is proportional to how often you ask, and Apple has published the meter for it inside the app being metered.
What it costs to be the answer
The bill on the measured Mac, for the current instance, alive 4 days and 14 hours: 5 minutes 57 seconds of processor time, about a minute and a quarter a day. Two honest footnotes belong on that number. First, this Mac runs statistics polling automation all day long, so its sysmond answers far more questions than a typical one will. Second, the quiet stretches are genuinely quiet: 0.05 seconds of CPU across three and a half idle mid-evening minutes, and zero log lines in the last six hours, because a server that talks numbers does not talk about itself.
Its memory tells the same story from another angle. Resident size fell from 3.4 MB to 1.3 MB between two checks four minutes apart, with no restart in between, the pressured exit machinery trimming pages nobody was reading. The virtual size, 466 GB of reserved address space, is the standard 64-bit mapping habit and means nothing on its own; the meaningful ceiling is the soft 30 MB jetsam limit launchd wrote over it, which makes the idea of this daemon ever becoming a memory problem someone else’s problem, by design.
See for yourself
Every fact above came from read only commands:
ps -Awwo pid,user,%cpu,rss,vsz,time,comm | grep [s]ysmond
launchctl print system/com.apple.sysmond
plutil -p /System/Library/LaunchDaemons/com.apple.sysmond.plist
man sysmond
otool -L "/System/Applications/Utilities/Activity Monitor.app/Contents/MacOS/Activity Monitor"
log show --predicate 'process == "sysmond"' --last 6h
The otool line is the one worth running first. Somewhere in that library list is /usr/lib/libsysmon.dylib, the client end of the conversation, and once you have seen it there, the row in Activity Monitor stops being a stranger. It is the clerk the app sends your questions to.
Questions
What is sysmond on my Mac?
It is the system monitor daemon, the root process that collects the per-process statistics apps query when they want to know what everything on the machine is doing. Activity Monitor is its best known client: its binary loads the client library for the com.apple.sysmond Mach service, and the daemon's own strings carry the app's column vocabulary, object classes named OS_sysmond_process and OS_sysmond_row, and the Disk tab's literal labels Bytes Read, Bytes Write, Operations Read and Operations Write. It has a manual page, unusual for this series, and the manual page says exactly one thing: The sysmond command should not be run directly. Dated May 11 2010. The searches that find this page, rows like what is sysmond on mac and what is sysmond in activity monitor, are people who met the name inside the very app it feeds and found no introduction anywhere else.
Why is sysmond using high CPU?
Because something is asking it questions. The daemon is a server: it wakes when a client connects through its Mach door, samples the process table, hands over rows, and goes back to waiting. On the Mac measured here, 29 days of uptime, it had collected 5 minutes 57 seconds of processor time in the 4 day 14 hour life of its current instance, about a minute and a quarter a day, on a machine that runs statistics-polling automation all day, and 0.05 seconds across a quiet three and a half minute stretch mid-evening. If its CPU column climbs for an hour, the cause is a client: Activity Monitor left open with Update Frequency set to 1 second, a menu bar stats tool polling every second, or a monitoring agent doing the same. Apple's own Activity Monitor help makes the point without naming the daemon: Increasing the update frequency may affect your overall system performance. Close the asker, not the answerer.
Why does sysmond show in Activity Monitor at all?
Because Activity Monitor lists every process on the machine, including the process producing the list. When you open the app, you start a subscription: fresh rows every few seconds, by default every 5. Each refresh is a round trip, app to daemon to kernel tables and back, and the app even offers to show you the subscription's cost. Its help says to watch the CPU Time column in the CPU pane as you change the update frequency, which is a rare thing, a settings pane asking you to measure the measurer. The two processes whose numbers move most when you do that experiment are Activity Monitor itself and sysmond. The observer effect, priced per second, inside the app's own documentation.
Is sysmond a virus or malware?
No. The real binary sits on the sealed system volume at /usr/libexec/sysmond, 273,568 bytes, a universal build that still carries x86_64 slices alongside arm64e, signed with the standard first party ladder, com.apple.sysmond, platform identifier 26. There are no virus or malware rows in its search family at all, which fits a job whose scariest symptom is a busy column in a monitoring app. The honest test is the one this series always falls back on: read the path. A process calling itself sysmond from a downloads folder or an applications folder would be a costume. The real one lives in /usr/libexec and nowhere else.
Can I quit or disable sysmond?
There is no supported way and nothing to gain. Its launch file has no KeepAlive at all: launchd starts it only when a client knocks on its Mach door, holds it for a minimum runtime of 10 seconds, and lets the system reclaim its memory the moment it is idle, which the launchd report records as four runs and a last exit reason of JETSAM_REASON_MEMORY_IDLE_EXIT, the deliberate kind. Quitting it in Activity Monitor buys you seconds of quiet before the next statistics question respawns it, and the app you quit it from is usually the client doing the asking. It also has no preferences domain, nothing in defaults, no settings of any kind: there is no knob, because there is nothing to tune. The levers are on the client side, the update frequency, and how many monitoring tools you run at once.
Why does sysmond use 466 GB of virtual memory?
It does not, and the number is not a problem. ps reports a virtual size of about 466 GB against a resident memory of 3.4 MB that later fell to 1.3 MB between two checks four minutes apart, without the process restarting. The virtual figure is reserved address space, mostly mappings that will never be touched, and it is the normal shape of a 64-bit macOS daemon, not a leak. What is worth knowing is the ceiling launchd has written over it: a soft jetsam memory limit of 30 MB, active and inactive, plus a thread limit of 32. If sysmond ever tries to hold more than that, the system reclaims it on purpose and starts a fresh one on the next question. The memory story is managed three layers deep before you would ever need to care.
Ready? Buy Ribbon for $15, or read all the guides.