For the TCP throughput tests, we use iperf3's -O "omit" option which
ignores results for the given time at the beginning of the test. Currently
we calculate this as 1/6th of the test measurement time. The purpose of
-O, however, is to skip over the TCP slow start period, which in no way
depends on the overall length of the test.
The slow start time is roughly speaking
log_2 ( max_window_size / MSS ) * round_trip_time
These factors all vary between tests and machines we're running on, but we
can estimate some reasonable bounds for them:
* The maximum window size is bounded by the buffer sizes at each end,
which shouldn't exceed 16MiB
* The mss varies with the MTU we use, but the smallest we use in tests is
~256 bytes
* Round trip time will vary with the system, but with these essentially
local transfers it will typically be well under 1ms (on my laptop it is
closer to 0.03ms)
That gives a worst case slow start time of about 16ms. Setting an omit
time of 0.1s uniformly is therefore more than enough, and substantially
smaller than what we calculate now for the default case (10s / 6 ~= 1.7s).
This reduces total time for the standard benchmark run by around 30s.
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
We always set --pacing-timer when invoking iperf3. However, the iperf3
man page implies this is only relevant for the -b option. We only use the
-b option for the UDP tests, not TCP, so remove --pacing-timer from the TCP
cases.
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
The TCP packet size used on the passt L2 link (qemu socket) makes a huge
difference to passt/pasta throughput; many of passt's overheads (chiefly
syscalls) are per-packet.
That packet size is largely determined by the MTU on the L2 link, so we
benchmark for a number of different MTUs. That works well for the guest to
host transfers. For the host to guest transfers, we purport to test for
different MTUs, but we're not actually adjusting anything interesting.
The host to guest transfers adjust the MTU on the "host's" (actually ns)
loopback interface. However, that only affects the packet size for the
socket going to passt, not the packet size for the L2 link that passt
manages - passt can and will repack the stream into packets of its own
size. Since the depacketization on that socket is handled by the kernel it
doesn't have a lot of bearing on passt's performance.
We can't fix this by changing the L2 link MTU from the guest side (as we do
for guest to host), because that would only change the guest's view of the
MTU, passt would still think it has the large MTU. We could test this by
using the --mtu option to passt, but that would require restarting passt
for each run, which is awkward in the current setup. So, for now, drop all
the "small MTU" tests for host to guest.
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
Packet size can make a big difference to UDP throughput, so it makes sense
to measure it for a variety of different sizes. Currently we do this by
adjusting the MTU on the relevant interface before running iperf3.
However, the UDP packet size has no inherent connection to the MTU - it's
controlled by the sender, and the MTU just affects whether the packet will
make it through or be fragmented. The only reason adjusting the MTU works
is because iperf3 bases its default packet size on the (path) MTU.
We can test this more simply by using the -l option to the iperf3 client
to directly control the packet size, instead of adjusting the MTU.
As well as simplifying this lets us test different packet sizes for host to
ns traffic. We couldn't do that previously because we don't have
permission to change the MTU on the host.
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
Currently we make TCP throughput measurements for spliced connections with
a number of different MTU values. However, the results from this aren't
really interesting.
Unlike with tap connections, spliced connections only involve the loopback
interface on host and container, not a "real" external interface. lo
typically has an MTU of 65535 and there is very little reason to ever
change that. So, the measurements for smaller MTUs are rarely going to be
relevant.
In addition, the fact that we can offload all the {de,}packetization to the
kernel with splice(2) means that the throughput difference between these
MTUs isn't very great anyway.
Remove the short MTUs and only show spliced throughput for the normal
65535 byte loopback MTU. This reduces runtime of the performance tests on
my laptop by about 1 minute (out of ~24 minutes).
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
Currently we start both the iperf3 server(s) and client(s) afresh each time
we want to make a bandwidth measurement. That's not really necessary as
usually a whole batch of bandwidth measurements can use the same server.
Split up the iperf3 directive into 3 directives: iperf3s to start the
server, iperf3 to make a measurement and iperf3k to kill the server, so
that we can start the server less often. This - and more importantly, the
reduced number of waits for the server to be ready - reduces runtime of the
performance tests on my laptop by about 4m (out of ~28minutes).
For now we still restart the server between IPv4 and IPv6 tests. That's
because in some cases the latency measurements we make in between use the
same ports.
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
Some older revisions used separate iperf3c and iperf3s test directives to
invoke the iperf3 client and server. Those were combined into a single
iperf3 directive some time ago, but a couple of places still have the old
syntax.
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
In practical terms, passt doesn't benefit from the additional
protection offered by the AGPL over the GPL, because it's not
suitable to be executed over a computer network.
Further, restricting the distribution under the version 3 of the GPL
wouldn't provide any practical advantage either, as long as the passt
codebase is concerned, and might cause unnecessary compatibility
dilemmas.
Change licensing terms to the GNU General Public License Version 2,
or any later version, with written permission from all current and
past contributors, namely: myself, David Gibson, Laine Stump, Andrea
Bolognani, Paul Holzinger, Richard W.M. Jones, Chris Kuhn, Florian
Weimer, Giuseppe Scrivano, Stefan Hajnoczi, and Vasiliy Ulyanov.
Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
Similarly to UDP cases, these were missing as it wasn't clear, when
the other tests were introduced, if using the global address of a
namespace, from the host, should have resulted in connections being
routed via the tap interface.
Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
Reviewed-by: David Gibson <david@gibson.dropbear.id.au>
These were missing as it wasn't clear, when the other tests were
introduced, if using the global address of a namespace, from the
host, should have resulted in traffic being routed via the tap
interface (as opposed to the loopback interface). We now clarified
that's actually the case.
Use same values and thresholds as the tests for loopback traffic, as
throughput figures currently indicate there isn't much difference.
Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
Reviewed-by: David Gibson <david@gibson.dropbear.id.au>
Now that we require 13c6be96618c ("net: stream: add unix socket")
in qemu to run the tests, we can also assume that commit df8d07081718
("virtio-net: fix bottom-half packet TX on asynchronous completion")
is present, as it was merged before that one.
This fixes the issue we attempted to work around in passt TCP and
UDP performance tests: finally drop that stuff.
Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
Starting tcp_rr, tcp_crr, udp_rr servers in the guest takes a bit
longer than starting the corresponding clients on the host, and we
end up starting clients before servers unless we add a delay there.
Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
It appears that if we run throughput tests with one-second periodic
reports, the sending side of the vhost channel used for SSH-based
command dispatch occasionally stops working altogether. I haven't
investigated this further, all I see is that output is truncated
at some point, and doesn't resume.
If we use gzip compression (ssh -C) this happens less frequently,
but it still happens, seemingly indicating the issue is probably
related to vhost itself.
Disable periodic reports in iperf3 clients. The -i options were
actually redundant, so remove them from both test files as well as
from test_iperf3().
Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
It looks like the workaround for the virtio_net TX hang issue is
working less reliably with the new command dispatch mechanism, I'm
not sure why. Switch to 10 seconds, at least for the moment.
Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
Reviewed-by: David Gibson <david@gibson.dropbear.id.au>
Otherwise, we're depending on having /sbin in $PATH. For some reason
I didn't completely grasp, with the new command dispatch mechanism
that's not the case anymore, even if I have /sbin in $PATH in the
parent shell.
Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
Reviewed-by: David Gibson <david@gibson.dropbear.id.au>
test_iperf3() is a pretty inscrutable mess of nested background processes.
It has a number of ugly sleeps needed to wait for things to complete.
Rewrite it to be cleaner:
* Use the construct (a & b & wait) to run 'a' and 'b' in parallel, but
then wait for them both to complete before continuing
* This allows us to wait for both the server and client to finish, rather
than sleeping
* Use jq to do all the math we need to get the final result, rather than
jq followed by some complicated 'bc' mangling
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Currently all the throughput tests are run for 30s. This is reflected in
both the actual parameters given to the iperf commands, but also in the
matching sleeps in test_iperf3.
Allow this to be adjusted more easily with a new parameter to test_iperf3.
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
[sbrivio: Reflect new parameter in comment to test_iperf3()]
Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
These two commands in the DSL to run an iperf client and server are always
used together, and some of the parameters must match between them. The
iperf3s must also be run more or less immediately after iperf3c, since
iperf3c will run a client in the background after a sleep and requires a
server to be running before it will work.
A bunch of things can be made cleaner if we make a single DSL command that
runs both sides of the test. For now make the combined command work
exactly like the two commands together did, warts and all.
This does lose the ability for the DSL scripts to give additional options
to the iperf3 server, but we weren't using that anyway.
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Currently test/run uses wildcards to run all of the tests in a directory.
However, that wildcard list is filtered down by the "onlyfor" directives
in the test files... usually to a single file.
Therefore, just explicitly list the files we *really* want to run for this
test mode. This makes it easier to see at the top level what tests will
be executed, and to change that list temporarily while debugging specific
failures.
This means the "onlyfor" directive no longer has any purpose, and we can
remove it. "onlyfor" was also the only used of the $MODE variable, so we
can remove that too.
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
David reports that dhclient-script(8) on Fedora needs a number of
binaries that are not included in PROGS of the current mbuto profile,
and we would also need to include hostnamectl(1) there, which will
fail without a systemd init.
Embed a minimal script for dhclient(8) in the profile itself, written
to /sbin/dhclient-script at boot, to just check what we need to check
out of DHCP and DHCPv6 functionality.
While at it, drop busybox and logger from PROGS, as we don't need them,
and add hostname(1). While DHCP option 12 isn't supported yet by the
DHCP implementation in passt, we should probably add it soon.
Note: owing to the simplicity of this script, we now need to bring up
the interface before starting dhclient: add this in test scripts where
it's not the case yet.
Reported-by: David Gibson <david@gibson.dropbear.id.au>
Suggested-by: David Gibson <david@gibson.dropbear.id.au>
Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
note that we need to bring up the interface before starting dhclient
A number of tests and examples use dhclient in both IPv4 and IPv6 modes.
We use "dhclient -6" for IPv6, but usually just "dhclient" for IPv4. Add
an explicit "-4" argument to make it more clear and explicit.
In addition, when dhclient is run from within pasta it usually won't be
"real" root, and so will not have access to write the default global pid
file. This results in a mostly harmless but irritating error:
Can't create /var/run/dhclient.pid: Permission denied
We can avoid that by using the --no-pid flag to dhclient.
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
ip(8)'s ability to take abbreviated arguments (e.g. "li sh" instead of
"link show") is very handy when using it interactively, but it doesn't make
for very readable scripts and examples when shown that way.
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
There are a few occurrences of this assignment, which are needed to
re-add ::1 as loopback address after the MTU has been increased
back from a value below 1280 bytes.
This one, however, is redundant, and causes an error in the
execution.
Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
I didn't have time to investigate the root cause for the virtio_net
TX hang yet. Add a quick work-around for the moment being.
Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
That test fails sometimes, it looks like iperf3 is still sending
initial messages that are too big. I'll need to figure out why,
but given that 256 bytes is not really an expected MTU, drop the
thresholds to zero for the moment being.
Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
With a recent 5.15 kernel, passing a huge window size to iperf3 with
lower MTUs makes iperf3 stop sending packets after a few seconds --
I haven't investigated this in detail, but the window size will be
adjusted dynamically anyway and not passing it doesn't actually
affect throughput, so simply drop the option.
Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
The throughput results in this test look quite variable, slightly
lower figures look reasonable anyway.
Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
With recent improvements, we're not CPU-bound at all while testing
UDP performance. Give the VM more memory and CPUs, forward two
additional ports, start up to four threads in parallel, and give
single iperf3 threads higher bandwidth targets.
Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
On most recent CPUs, that's a better indication of all-core turbo
frequency, or non-turbo frequency, than /proc/cpuinfo.
Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
