Change default RDMA settings to 128x 129K buffers

129K to leave extra space for the header

The problem with 8x 1M buffers is that the following happens with,
for example, 2 OSDs and 4M T1Q1 write:
- Server posts 8 receives
- Client posts 8 sends
- WRs are processed by the RDMA stack, but the OSD doesn't have the time
  to handle them and doesn't refill buffers
- Client posts 1 more send
- RNR retransmission happens and performance drops to zero

Overall it seems that RDMA support should be reworked to use real 'RDMA'
operations i.e. operations writing into remote memory. This has an
additional advantage of avoiding a copy at the receive side of the OSD.
This commit is contained in:
Vitaliy Filippov
2021-11-21 12:05:52 +03:00
parent f0ebfae3b8
commit 660c3f7b0d
2 changed files with 4 additions and 4 deletions
+3 -3
View File
@@ -147,13 +147,13 @@ void osd_messenger_t::parse_config(const json11::Json & config)
this->rdma_max_sge = 128;
this->rdma_max_send = config["rdma_max_send"].uint64_value();
if (!this->rdma_max_send)
this->rdma_max_send = 32;
this->rdma_max_send = 1;
this->rdma_max_recv = config["rdma_max_recv"].uint64_value();
if (!this->rdma_max_recv)
this->rdma_max_recv = 8;
this->rdma_max_recv = 128;
this->rdma_max_msg = config["rdma_max_msg"].uint64_value();
if (!this->rdma_max_msg || this->rdma_max_msg > 128*1024*1024)
this->rdma_max_msg = 1024*1024;
this->rdma_max_msg = 129*1024;
#endif
this->receive_buffer_size = (uint32_t)config["tcp_header_buffer_size"].uint64_value();
if (!this->receive_buffer_size || this->receive_buffer_size > 1024*1024*1024)