cisco-system-traffic-generator/trex-core

jitter got 0 sometimes

zzsuki opened this issue · 2 comments

zzsuki commented

Question

  1. my stream is created by this Class
class StreamUtil:
    UDP_HEADER_SIZE = 42

    def create_latency_stream(self, src_ip, dst_ip, src_port, dst_port, payload, pg_id=1024):
        return STLStream(
            packet=STLPktBuilder(
                pkt=Ether() / IP(src=src_ip, dst=dst_ip) / UDP(sport=src_port, dport=dst_port) / payload
            ),
            mode=STLTXCont(pps=1000),
            flow_stats=STLFlowLatencyStats(pg_id=pg_id)
        )

    def calculate_udp_payload(self, size):
        # eth：6(dst mac) + 6(src mac) + 2(type)
        # ip: 4(src ip) + 4(dst ip) + 12(other arguments)
        # udp: 2(src port) + 2(dst port) + 4(length + checksum) + payload
        # total non data part = 14 + 20 + 8 = 42
        # so payload should be `packet_size - 42`
        return size - self.UDP_HEADER_SIZE

  1. When I run script based on this and useget_pgid_stats to get jitter 20times every 20s, it gets some 0 record (sometime even all records return 0) . I'm not sure if this is caused by calculating precision?

  2. I have some confusion about STLTXCont(pps=1000) which number is proper for pps, I read from official docs and realized it shouldn't increase too big; when i test this without DUT（connect two PORTS directly by cable）it still has almost 200us latency. Is latency has some relationship with this pps?

hhaim commented

please use our mailers to questions

zzsuki commented