rlc: increase RLC Tx SDU queue

the Tx SDU has been set to 128 PDUs because this gives approx.
150kByte UL buffer state (max value for Rel 8 BSR).

Previously, however, we've always reported too much pending UL
data because we weren't subtracting the transmitted data from
the reported data.

with this newer BSR reporting, however, even with full buffers we would
never report full buffers because the queues size is too short.

This seems to cause issues for iperf TCP sessions not resulting
in full DL rate, see isse #1703. It seems that the TCP bandwitdh
probing works better with larger buffers (in our case this can
only be achieved with more PDUs in the pool).

This should fix #1703
This commit is contained in:
Andre Puschmann 2020-09-08 15:53:59 +02:00
parent 569d8ee860
commit 6a56476b2a
1 changed files with 1 additions and 1 deletions

View File

@ -124,7 +124,7 @@ struct rlc_um_nr_config_t {
int32_t t_reassembly_ms; // Timer used by rx to detect PDU loss (ms)
};
#define RLC_TX_QUEUE_LEN (128)
#define RLC_TX_QUEUE_LEN (256)
enum class srslte_rat_t { lte, nr, nulltype };
inline std::string to_string(const srslte_rat_t& type)