> On Jul 25, 2017, at 4:04 AM, Paul Tsvika <[email protected]> wrote: > > Hi > > I've some updates and added inline text in my post. > > > Thanks. > > >> >> Memory: 32G ( 16 G on two channels ) >> CPU: Xeon-D 1557 >> NIC: Intel X552/X557 10GBASE-T >> >> CPU layout: ( Graph generated from cpu_layout.py ) >> >> Core and Socket Information (as reported by '/proc/cpuinfo') >> ============================================================ >> >> >> >> >> >> >> >> >> >> >> >> >> >> >> >> >> >> *cores = [0, 1, 2, 3, 4, 5, 8, 9, 10, 11, 12, 13]sockets = [0] >> Socket 0 -------- Core 0 [0, 12] Core 1 [1, >> 13] Core 2 [2, 14] Core 3 [3, 15] Core 4 [4, >> 16] Core 5 [5, 17] Core 8 [6, 18] Core 9 [7, >> 19] Core 10 [8, 20] Core 11 [9, 21] Core 12 [10, >> 22] Core 13 [11, 23]* >> ============================================================= >> >> My motherboard has 2 10G ports so that I set up the testing experiment in >> loopback. >> >> I have compiled DPDK successfully and below is the steps how I configured >> pktgen >> >> $ sudo modprobe uio >> $ sudo insmod ./build/kmod/igb_uio.ko >> $ sudo ./usertools/dpdkdevbind.py -b igb_uio xxx:xx.0 xxx:xx.1 >> $ sudo ./app/pktgen -c 0x1f -n 2 -- -P -m "[1:3].0, [2:4].1" >> <pktgen> start 0 >> >> I use core 1 and core 3 work for port 0 tx/rx respectively. Core 2 and >> core 4 work for port 1 tx/rx respectively. >> >> I have some questions listed below: >> >> 1. >> >> I checked other developers who run pktgen successfully should get the >> TX/RX rate almost the same and the value should hit 10000 or 99xx. >> I have not figured out yet why the rate went down when I run pktgen. >> >> >> >> >> >> * Flags:Port : P--------------:0 P--------------:1 >> 10318718/0Link State : <UP-10000-FD> <UP-10000-FD> >> ----TotalRate----Pkts/s Max/Rx : 0/0 >> 10318848/8096382 10318848/8096382 Max/Tx : >> 14880683/11528400 0/0 14880683/11528400MBits/s >> Rx/Tx : 0/7747 5440/0 5440/7747 >> * >> > *[PT 1:] Any developer has the experience to this ? *
The only other way you can get a performance difference is the NIC’s are in PCIe slots which are not able to handle the performance. Did you check if the slots are both 8xPCIe slots. > >> >> >> >> >> >> >> >> >> >> >> >> >> >> >> >> >> >> >> >> >> >> >> >> >> >> *Broadcast : 0 >> 0Multicast : 0 0 64 >> Bytes : 0 4603033247 65-127 >> : 0 0 128-255 >> : 0 0 256-511 >> : 0 0 512-1023 >> : 0 0 1024-1518 >> : 0 0Runts/Jumbos >> : 0/0 0/0Errors Rx/Tx >> : 0/0 0/0Total Rx Pkts >> : 0 4596736394 Tx Pkts : >> 6588970508 0 Rx MBs : >> 0 3089006 Tx MBs : >> 4427788 0ARP/ICMP Pkts : >> 0/0 0/0Tx Count/% Rate : Forever /100% >> Forever /100%Pattern Type : abcd... abcd...Tx >> Count/% Rate : Forever /100% Forever /100%PktSize/Tx Burst >> : 64 / 32 64 / 32 -------------------Src/Dest >> Port : 1234 / 5678 1234 / 5678sPkt Type:VLAN ID : >> IPv4 / TCP:0001 IPv4 / TCP:0001Dst IP Address : >> 192.168.1.1 192.168.0.1Src IP Address : 192.168.0.1/24 >> <http://192.168.0.1/24> 192.168.1.1/24 <http://192.168.1.1/24> >> -------------------Dst MAC Address : 0c:c4:7a:c5:43:7b >> 0c:c4:7a:c5:43:7a-- Pktgen Ver: 3.3: 8086:15ad/03:00.0 >> 8086:15ad/03:00.1 -------------------* >> >> >> 2.1 >> >> My start_log below: >> >> Copyright (c) <2010-2017>, Intel Corporation. All rights reserved. >> Pktgen created by: Keith Wiles -- >>> Powered by Intel® DPDK <<< >> >>>>> Packet Burst 32, RX Desc 512, TX Desc 1024, mbufs/port 8192, mbuf >> cache 1024 >> >> *[PT]: looks like packet size is 32, should it suppose to be 64? Can I >> configure it?* >> *Also, why RX Desc and TX Desc is not the same ?* >> >> === port to lcore mapping table (# lcores 5) === >> lcore: 0 1 2 3 4 Total >> port 0: ( D: T) ( 1: 0) ( 0: 0) ( 0: 1) ( 0: 0) = ( 1: 1) >> port 1: ( D: T) ( 0: 0) ( 1: 0) ( 0: 0) ( 0: 1) = ( 1: 1) >> Total : ( 0: 0) ( 1: 0) ( 1: 0) ( 0: 1) ( 0: 1) >> Display and Timer on lcore 0, rx:tx counts per port/lcore >> >> Configuring 2 ports, MBUF Size 2176, MBUF Cache Size 1024 >> Lcore: >> 1, RX-Only >> RX_cnt( 1): (pid= 0:qid= 0) >> 2, RX-Only >> RX_cnt( 1): (pid= 1:qid= 0) >> 3, TX-Only >> TX_cnt( 1): (pid= 0:qid= 0) >> 4, TX-Only >> TX_cnt( 1): (pid= 1:qid= 0) >> >> Port : >> 0, nb_lcores 2, private 0x986ec0, lcores: 1 3 >> 1, nb_lcores 2, private 0x989220, lcores: 2 4 >> >> >> >> ** Default Info (0000:03:00.0, if_index:0) ** >> max_vfs : 0, min_rx_bufsize :1024, max_rx_pktlen : 15872 >> max_rx_queues : 128, max_tx_queues : 64 >> max_mac_addrs : 128, max_hash_mac_addrs:4096, max_vmdq_pools: 64 >> rx_offload_capa: 79, tx_offload_capa : 191, reta_size : 512, >> flow_type_rss_offloads:0000000000038d34 >> vmdq_queue_base: 0, vmdq_queue_num : 128, vmdq_pool_base: 0 >> ** RX Conf ** >> pthresh : 8, hthresh : 8, wthresh : 0 >> Free Thresh : 32, Drop Enable : 0, Deferred Start : 0 >> ** TX Conf ** >> pthresh : 32, hthresh : 0, wthresh : 0 >> Free Thresh : 32, RS Thresh : 32, Deferred Start : 0, >> TXQ Flags:00000f01 >> >> Create: Default RX 0:0 - Memory used (MBUFs 8192 x (size 2176 + Hdr >> 128)) + 192 = 18433 KB headroom 128 2176 >> Set RX queue stats mapping pid 0, q 0, lcore 1 >> >> >> Create: Default TX 0:0 - Memory used (MBUFs 8192 x (size 2176 + Hdr >> 128)) + 192 = 18433 KB headroom 128 2176 >> Create: Range TX 0:0 - Memory used (MBUFs 8192 x (size 2176 + Hdr >> 128)) + 192 = 18433 KB headroom 128 2176 >> Create: Sequence TX 0:0 - Memory used (MBUFs 8192 x (size 2176 + Hdr >> 128)) + 192 = 18433 KB headroom 128 2176 >> Create: Special TX 0:0 - Memory used (MBUFs 64 x (size 2176 + Hdr >> 128)) + 192 = 145 KB headroom 128 2176 >> >> >> Port memory used = 73873 KB >> Initialize Port 0 -- TxQ 1, RxQ 1, Src MAC 0c:c4:7a:c5:43:7a >> >> ** Default Info (0000:03:00.1, if_index:0) ** >> max_vfs : 0, min_rx_bufsize :1024, max_rx_pktlen : 15872 >> max_rx_queues : 128, max_tx_queues : 64 >> max_mac_addrs : 128, max_hash_mac_addrs:4096, max_vmdq_pools: 64 >> rx_offload_capa: 79, tx_offload_capa : 191, reta_size : 512, >> flow_type_rss_offloads:0000000000038d34 >> vmdq_queue_base: 0, vmdq_queue_num : 128, vmdq_pool_base: 0 >> ** RX Conf ** >> pthresh : 8, hthresh : 8, wthresh : 0 >> Free Thresh : 32, Drop Enable : 0, Deferred Start : 0 >> ** TX Conf ** >> pthresh : 32, hthresh : 0, wthresh : 0 >> Free Thresh : 32, RS Thresh : 32, Deferred Start : 0, >> TXQ Flags:00000f01 >> >> Create: Default RX 1:0 - Memory used (MBUFs 8192 x (size 2176 + Hdr >> 128)) + 192 = 18433 KB headroom 128 2176 >> Set RX queue stats mapping pid 1, q 0, lcore 2 >> >> >> Create: Default TX 1:0 - Memory used (MBUFs 8192 x (size 2176 + Hdr >> 128)) + 192 = 18433 KB headroom 128 2176 >> Create: Range TX 1:0 - Memory used (MBUFs 8192 x (size 2176 + Hdr >> 128)) + 192 = 18433 KB headroom 128 2176 >> Create: Sequence TX 1:0 - Memory used (MBUFs 8192 x (size 2176 + Hdr >> 128)) + 192 = 18433 KB headroom 128 2176 >> Create: Special TX 1:0 - Memory used (MBUFs 64 x (size 2176 + Hdr >> 128)) + 192 = 145 KB headroom 128 2176 >> >> >> Port memory used = 73873 KB >> Initialize Port 1 -- TxQ 1, RxQ 1, Src MAC 0c:c4:7a:c5:43:7b >> >> Total memory used = 147746 KB >> Port 0: Link Down <Enable promiscuous mode> >> *[PT]: I don't know why the Port ) went down at this point. Is there any >> other thing I should check ?* >> > * [PT 2]: After updating the NIC firmware, the drop issue has been > fixed. * > >> >> >> Port 1: Link Up - speed 10000 Mbps - full-duplex <Enable promiscuous mode> >> >> pktgen_packet_capture_init: mz->len 4194304 >> >> === Display processing on lcore 0 >> RX processing lcore: 1 rx: 1 tx: 0 >> For RX found 1 port(s) for lcore 1 >> RX processing lcore: 2 rx: 1 tx: 0 >> For RX found 1 port(s) for lcore 2 >> TX processing lcore: 3 rx: 0 tx: 1 >> For TX found 1 port(s) for lcore 3 >> TX processing lcore: 4 rx: 0 tx: 1 >> For TX found 1 port(s) for lcore 4 >> >> >> >> >> >> Thanks for any feedback. >> >> >> -- >> P.T >> > > > > -- > P.T Regards, Keith
