| Commit message (Collapse) | Author | Age | Files | Lines |
|---|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
The ECN marks were the only source of congestion information, and
going into congestion collapse (dropping a lot of ECN packets and
causing loss of signal) was unrecoverable.
The sender's slow-start ramp clock now tracks a measured RTT to avoid
overshooting into congestion collapse on the previously fixed (14ms)
doubling on long-delay (e.g. 100ms) links.
The flow allocator now carries a periodic heartbeat/ack to keep RTT
estimates up-to-date. The heartbeat also serves as a congestion
collapse detector: consecutive heartbeat losses will go into loss
recovery, halving the window and pausing Additive Increase until the
path is clear. The DT component now has a max_rtt config parameter that
serves as a hint for the layer RTT to optimize slow-start for
low-latency networks.
These two mechanisms follow TCP, however only during slow start up to
the first ECN signal and to recover from a total collapse until the
ECN signal returns. MB-ECN is still fully RTT-independent in the AI/PD
region. A shorter RTT path will just reach its equal fair share faster
than a longer RTT path.
Signed-off-by: Dimitri Staessens <dimitri@ouroboros.rocks>
Signed-off-by: Sander Vrijders <sander@ouroboros.rocks>
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Congestion avoidance is a property of the layer, orthogonal to ARQ
and to flow control: FRCP retransmits and lets the peer pace the
sender, per flow, end-to-end; the IPCP paces path aggregates. Each
signal means one thing: a loss triggers a retransmission, a mark
means congestion, the peer window means a slow receiver. Every flow
is paced by the same rate law whatever its QoS, so a greedy raw
sender shares a bottleneck fairly with a reliable stream.
The unit of control is the (destination address, QoS cube)
aggregate: all flows toward that destination share one controller
and one rate; a start-time fair-queuing pacer divides the rate
across them by deadline instead of blocking the send path, and a
new flow rides the aggregate's estimates at its current rate, with
no probing of its own. Slow start runs once per aggregate.
The congestion signal is a multi-bit magnitude: forwarders mark
packets with their standing queue depth, MAX-combined across hops,
so a packet carries the deepest queue on its path. The receiver
feeds back a time-integral mean over a window that adapts to the
flow's byte rate, measuring a slow flow with the same fidelity as a
fast one. The sender runs AIMD scaled by elapsed wall-clock time,
which makes the steady-state allocation RTT-independent.
The PCI gains one byte: the path capacity as a quarter-log2 code.
Forwarders estimate their egress rate from busy-period drain and
MIN-stamp the byte, the receiver returns the window minimum with
its feedback, and the sender scales its rate floor and additive
slope to the bottleneck (C / 32). A deep cut implies a backlogged
bottleneck and a backlogged bottleneck advertises its capacity, so
the scaled floor is live exactly when recovery needs it: the probe
heals a halving in seconds at any link rate, and the floor bounds
the deepest hole to a factor 32 below the bottleneck.
Signed-off-by: Dimitri Staessens <dimitri@ouroboros.rocks>
Signed-off-by: Sander Vrijders <sander@ouroboros.rocks>
|
| |
|
|
|
| |
Signed-off-by: Dimitri Staessens <dimitri@ouroboros.rocks>
Signed-off-by: Sander Vrijders <sander@ouroboros.rocks>
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
This is a full revision of the IRMd internal implementation.
The registry is now a proper subcomponent managing its own internal
lock (a single mutex). Some tests are added for the registry and its
data structures. Some macros for tests are added in <ouroboros/test.h>.
Flow allocation is now more symmetric between the client side (alloc)
and server size (accept). Each will create a flow in pending state
(ALLOC_PENDING/ACCEPT_PENDING) that is potentially fulfilled by an
IPCP using respond_alloc and respond_accept primitives. Deallocation
is split in flow_dealloc (application side) and ipcp_flow_dealloc
(IPCP side) to get the flow in DEALLOC_PENDING and DEALLOCATED state.
Cleanup of failed flow allocation is now properly handled instead of
relying on the sanitizer thread. The new sanitizer only needs to
monitor crashed processes.
On shutdown, the IRMd will now detect hanging processes and SIGKILL
them and clean up their fuse mountpoints if needed.
A lot of other things have been cleaned up and shuffled around a bit.
Signed-off-by: Dimitri Staessens <dimitri@ouroboros.rocks>
Signed-off-by: Sander Vrijders <sander@ouroboros.rocks>
|
| |
|
|
|
|
|
| |
Slow but steady.
Signed-off-by: Dimitri Staessens <dimitri@ouroboros.rocks>
Signed-off-by: Sander Vrijders <sander@ouroboros.rocks>
|
| |
|
|
|
|
|
| |
2022 was a rather slow year...
Signed-off-by: Dimitri Staessens <dimitri@ouroboros.rocks>
Signed-off-by: Sander Vrijders <sander@ouroboros.rocks>
|
| |
|
|
|
|
|
| |
Growing pains.
Signed-off-by: Dimitri Staessens <dimitri@ouroboros.rocks>
Signed-off-by: Sander Vrijders <sander@ouroboros.rocks>
|
|
|
The policies were all in a single folder pol/, and have been moved to
a folder per component/mechanism to keep things a bit more orderly.
Signed-off-by: Dimitri Staessens <dimitri@ouroboros.rocks>
Signed-off-by: Sander Vrijders <sander@ouroboros.rocks>
|
