diff --git a/src/core/app.lua b/src/core/app.lua
index 80a4a9a29c..b997307586 100644
--- a/src/core/app.lua
+++ b/src/core/app.lua
@@ -76,6 +76,8 @@ busywait = false
 
 -- Profiling with vmprofile --------------------------------
 
+vmprofile_enabled = true
+
 -- Low-level FFI
 ffi.cdef[[
 int vmprofile_get_profile_size();
@@ -93,17 +95,21 @@ local function getvmprofile (name)
 end
 
 function setvmprofile (name)
-   C.vmprofile_set_profile(getvmprofile(name))
+   if vmprofile_enabled then
+      C.vmprofile_set_profile(getvmprofile(name))
+   end
 end
 
 function clearvmprofiles ()
-   jit.vmprofile.stop()
-   for name, profile in pairs(vmprofiles) do
-      shm.unmap(profile)
-      shm.unlink("vmprofile/"..name..".vmprofile")
-      vmprofiles[name] = nil
+   if vmprofile_enabled then
+      jit.vmprofile.stop()
+      for name, profile in pairs(vmprofiles) do
+         shm.unmap(profile)
+         shm.unlink("vmprofile/"..name..".vmprofile")
+         vmprofiles[name] = nil
+      end
+      jit.vmprofile.start()
    end
-   jit.vmprofile.start()
 end
 
 -- True when the engine is running the breathe loop.
diff --git a/src/lib/ptree/worker.lua b/src/lib/ptree/worker.lua
index 971213f230..3925fc642d 100644
--- a/src/lib/ptree/worker.lua
+++ b/src/lib/ptree/worker.lua
@@ -21,6 +21,7 @@ local worker_config_spec = {
    duration = {},
    measure_latency = {default=true},
    measure_memory = {default=true},
+   profile = {default=true},
    no_report = {default=false},
    report = {default={showapps=true,showlinks=true}},
    Hz = {default=1000},
@@ -46,6 +47,7 @@ function new_worker (conf)
    if conf.measure_memory then
       timer.activate(memory_info.HeapSizeMonitor.new():timer())
    end
+   engine.vmprofile_enabled = conf.profile
    return ret
 end
 
@@ -100,16 +102,12 @@ function Worker:handle_actions_from_manager()
 end
 
 function Worker:main ()
-   local vmprofile = require("jit.vmprofile")
    local stop = engine.now() + self.duration
    local next_time = engine.now()
 
-   -- Setup vmprofile.
-   engine.setvmprofile("engine")
-   vmprofile.start()
-
    if not engine.auditlog_enabled then engine.enable_auditlog() end
 
+   engine.setvmprofile("engine")
    repeat
       self.breathe()
       if next_time < engine.now() then
diff --git a/src/lib/scheduling.lua b/src/lib/scheduling.lua
index ab3ac2f0d8..2ddb9f3811 100644
--- a/src/lib/scheduling.lua
+++ b/src/lib/scheduling.lua
@@ -16,6 +16,7 @@ local scheduling_opts = {
    cpu = {},                  -- CPU index (integer).
    real_time = {},            -- Boolean.
    ingress_drop_monitor = {}, -- Action string: one of 'flush' or 'warn'.
+   profile = {default=true},  -- Boolean.
    busywait = {default=true}, -- Boolean.
    eval = {}                  -- String.
 }
@@ -42,6 +43,12 @@ function sched_apply.busywait (busywait)
    engine.busywait = busywait
 end
 
+function sched_apply.profile (profile)
+   engine.vmprofile_enabled = profile
+   local vmprofile = require('jit.vmprofile')
+   if profile then vmprofile.start() else vmprofile.stop() end
+end
+
 function sched_apply.eval (str)
    loadstring(str)()
 end
diff --git a/src/program/lwaftr/run/README b/src/program/lwaftr/run/README
index 62d0eef32e..534b20e581 100644
--- a/src/program/lwaftr/run/README
+++ b/src/program/lwaftr/run/README
@@ -49,11 +49,10 @@ Optional arguments:
 
 Optional arguments for debugging and profiling:
        -v                       Verbose (repeat for more verbosity).
+       --profile                Enable the low-overhead sampling
+                                profiler.
        -t FILE, --trace FILE    Record a trace of any run-time "snabb
                                 config" commands to FILE.
-       -jv, -jv=FILE            Print out when traces are recorded.
-       -jp, -jp=MODE,FILE       Profile the system by method.
-       -jtprof                  Profile the system by trace.
        -b FILENAME, --bench-file FILENAME
                                 Write any benchmarking data to FILENAME.
        -D SECONDS               Stop after SECONDS, for debugging
diff --git a/src/program/lwaftr/run/run.lua b/src/program/lwaftr/run/run.lua
index cdf11b9f72..5690aac2b1 100644
--- a/src/program/lwaftr/run/run.lua
+++ b/src/program/lwaftr/run/run.lua
@@ -50,7 +50,7 @@ function parse_args(args)
    local conf_file, v4, v6
    local ring_buffer_size
    local opts = { verbosity = 0 }
-   local scheduling = { ingress_drop_monitor = 'flush' }
+   local scheduling = { ingress_drop_monitor = 'flush', profile = false }
    local handlers = {}
    function handlers.n (arg) opts.name = assert(arg) end
    function handlers.v () opts.verbosity = opts.verbosity + 1 end
@@ -100,13 +100,13 @@ function parse_args(args)
                   .." (valid values: flush, warn, off)")
       end
    end
-   function handlers.j(arg) scheduling.j = arg end
+   function handlers.profile() scheduling.profile = true end
    function handlers.h() show_usage(0) end
-   lib.dogetopt(args, handlers, "b:c:vD:yhir:n:j:t:",
+   lib.dogetopt(args, handlers, "b:c:vD:yhir:n:t:",
      { conf = "c", name = "n", cpu = 1, v4 = 1, v6 = 1,
        ["on-a-stick"] = 1, virtio = "i", ["ring-buffer-size"] = "r",
        ["real-time"] = 0, mirror = 1, ["ingress-drop-monitor"] = 1,
-       verbose = "v", trace = "t", ["bench-file"] = "b",
+       verbose = "v", trace = "t", ["bench-file"] = "b", ["profile"] = 0,
        duration = "D", hydra = "y", help = "h" })
    if ring_buffer_size ~= nil then
       if opts.virtio_net then