网络设备测试艺术(2)–应用层测试
•
文章
挂表验收的流氓又回归了,这次带大家看看应用层的测试方法
网络设备进入到应用层以后,各种系统的缺陷是各个公司及其不愿暴露的,同时应用层的防火墙/流控等设备性能测试上,整个行业基本上都装傻,包乱序不管,指标按照物理接口大包转发带宽忽悠,实际并发连接数超高但新建连接数一坨屎.或者新建也很好,就是连接拆的慢,大象流处理有问题等等..
测试环境
依旧还是那个ASR1000，上一期测试过的ESP100引擎。这个基于多核心的处理器的平台牛x之处就在于，它不光是一台路由器，还支持IPSecVPN几十个G的吞吐，同时还支持100G的防火墙和NAT，以及几十G的流控和应用识别。想当年我可是拿它直接参加某运营商防火墙集采还中标的。
ASR1009X基本NAT配置
我们这次以NAT为例来测试，配置如下：
interface HundredGigE0/1/0
ip address 1.1.1.1 255.255.255.0
ip nat inside
!
interface HundredGigE1/1/0
ip address 2.2.2.1 255.255.255.0
ip nat outside
!
ip nat translation finrst-timeout 1
ip nat translation max-entries 10000000
ip nat pool AAA 17.0.0.1 17.0.0.254 prefix-length 24
ip nat inside source list 10 pool AAA overload
!
ip route 16.0.0.0 255.0.0.0 1.1.1.2
ip route 48.0.0.0 255.0.0.0 2.2.2.2
!
ip access-list standard 10
10 permit 16.0.0.0 0.255.255.255Trex平台配置
通常我们在进行流量测试时需要支持超过一千万会话的场景，因此对于很多平台也是一个考验，因此需要在Trex的配置文件中增加Memory
[root@Trex v2.86]新建连接测试作弊手段
通常的测试方法是只测试新建，期间不会拆掉连接，然后设备方通常说，顺便这个测试例把并发连接数测了
客户一定要要求测试新建和拆连接同时测试时的最大会话建立速度。
新建连接测试的配置如下,其中我们把Delay即会话长度缩短，使得整个系统建连接以后就立即拆掉
./t-rex-64-o -f astf/http_high_active_flows.py -m 200000 -t delay=1 --astf --no-ofed-check -c 20astf/http_high_active_flows.py的内容：
其中
-m 200000为新建连接数为200K每秒，流量建立起来以后的统计情况：-Per port stats table
ports | 0 | 1
-----------------------------------------------------------------------------------------
opackets | 6924918 | 6180533
obytes | 887754867 | 628870502
ipackets | 6179461 | 6180499
ibytes | 628765569 | 829747853
ierrors | 0 | 0
oerrors | 0 | 0
Tx Bw | 617.56 Mbps | 434.81 Mbps
-Global stats enabled
Cpu Utilization : 5.0 % 2.1 Gb/core
Platform_factor : 1.0
Total-Tx : 1.05 Gbps
Total-Rx : 1.01 Gbps
Total-PPS : 1.14 Mpps
Total-CPS : 199.23 Kcps
Expected-PPS : 0.00 pps
Expected-CPS : 0.00 cps
Expected-L7-BPS : 0.00 bps
Active-flows : 1827860 Clients : 0 Socket-util : 0.0000 %
Open-flows : 2291151 Servers : 0 Socket : 0 Socket/Clients : -nan
drop-rate : 0.00 bps
current time : 50.9 sec
test duration : 3549.1 sec在这个界面中点击
t可以看到具体的会话统计| client | server |
-----------------------------------------------------------------------------------------
m_active_flows | 1663411 | 1604109 | active open flows
m_est_flows | 1602288 | 1602276 | active established flows
m_tx_bw_l7_r | 398.33 Mbps | 190.36 Mbps | tx L7 bw acked
m_tx_bw_l7_total_r | 398.31 Mbps | 190.36 Mbps | tx L7 bw total
m_rx_bw_l7_r | 190.37 Mbps | 398.31 Mbps | rx L7 bw acked
m_tx_pps_r | 799.93 Kpps | 799.83 Kpps | tx pps
m_rx_pps_r | 799.84 Kpps | 999.79 Kpps | rx pps
m_avg_size | 46.00 B | 40.89 B | average pkt size
m_tx_ratio | 100.00 % | 100.00 % | Tx acked/sent ratio
- | --- | --- |
m_traffic_duration | 75.51 sec | 75.40 sec | measured traffic duration
- | --- | --- |
TCP | --- | --- |
- | --- | --- |
tcps_connattempt | 15078482 | 0 | connections initiated
tcps_accepts | 0 | 15018249 | connections accepted
tcps_connects | 15017359 | 15016416 | connections established
tcps_closed | 13415071 | 13414140 | conn. closed (includes drops)
tcps_segstimed | 45111355 | 45049260 | segs where we tried to get rtt
tcps_rttupdated | 45046569 | 45043726 | times we succeeded
tcps_sndtotal | 60125024 | 60065678 | total packets sent
tcps_sndpack | 15017359 | 15016416 | data packets sent
tcps_sndbyte | 3754542018 | 1786953504 | data bytes sent by application
tcps_sndbyte_ok | 3739322391 | 1786953504 | data bytes sent by tcp
tcps_sndctrl | 15078482 | 0 | control (SYN|FIN|RST) packets sent
tcps_sndacks | 30029183 | 45049262 | ack-only packets sent
tcps_rcvpack | 30029179 | 30031011 | packets received in sequence
tcps_rcvbyte | 1786846166 | 3739087584 | bytes received in sequence
tcps_rcvackpack | 30029210 | 45043726 | rcvd ack packets
tcps_rcvackbyte | 3738870705 | 1786736805 | tx bytes acked by rcvd acks
tcps_rcvackbyte_of | 15013665 | 30029131 | tx bytes acked by rcvd acks - overflow acked
tcps_preddat | 15015512 | 0 | times hdr predict ok for data
- | --- | --- |
UDP | --- | --- |
- | --- | --- |
- | --- | --- |
Flow Table | --- | --- |
- | --- | --- |
err_cwf | 40 | 0 | *client pkt without flow
redirect_rx_ok | 2 | 0 | redirect to rx OK当您打过载的时候，应该可以看到如下的错误，一方面是整个设备的并发连接数会因为太忙拆不掉而急剧上升，另一方面是
t统计页面会有错误日志,例如我们打到1M CPS时,请注意下表中带*的统计，表示各种错误就开始出现了。
| client | server |
-----------------------------------------------------------------------------------------
m_active_flows | 20971518 | 1109319 | active open flows
m_est_flows | 1065539 | 1070179 | active established flows
m_tx_bw_l7_r | 301.40 Mbps | 136.45 Mbps | tx L7 bw acked
m_tx_bw_l7_total_r | 325.77 Mbps | 144.00 Mbps | tx L7 bw total
m_rx_bw_l7_r | 144.04 Mbps | 301.31 Mbps | rx L7 bw acked
m_tx_pps_r | 1.48 Mpps | 609.74 Kpps | tx pps
m_rx_pps_r | 589.02 Kpps | 726.25 Kpps | rx pps
m_avg_size | 26.96 B | 40.96 B | average pkt size
m_tx_ratio | 92.52 % | 94.76 % | Tx acked/sent ratio
- | --- | --- |
m_traffic_duration | 58.09 sec | 57.98 sec | measured traffic duration
- | --- | --- |
TCP | --- | --- |
- | --- | --- |
tcps_connattempt | 27575473 | 0 | connections initiated
tcps_accepts | 0 | 7670600 | connections accepted
tcps_connects | 7669494 | 7631460 | connections established
tcps_closed | 6603955 | 6561281 | conn. closed (includes drops)
tcps_segstimed | 42875900 | 22903061 | segs where we tried to get rtt
tcps_rttupdated | 20041731 | 22790403 | times we succeeded
tcps_sndtotal | 78011576 | 30534522 | total packets sent
tcps_sndpack | 7669494 | 7631457 | data packets sent
tcps_sndbyte | 6866292777 | 908143383 | data bytes sent by application
tcps_sndbyte_ok | 1909704006 | 908143383 | data bytes sent by tcp
tcps_sndctrl | 55110893 | 1 | control (SYN|FIN|RST) packets sent
tcps_sndacks | 15231189 | 22903064 | ack-only packets sent
tcps_rcvpack | 15231189 | 15232464 | packets received in sequence
tcps_rcvbyte | 908081027 | 1900233540 | bytes received in sequence
tcps_rcvackpack | 15231190 | 22790403 | rcvd ack packets
tcps_rcvackbyte | 1900102566 | 904519476 | tx bytes acked by rcvd acks
tcps_rcvackbyte_of | 7600256 | 15189399 | tx bytes acked by rcvd acks - overflow acked
tcps_preddat | 7630933 | 0 | times hdr predict ok for data pkts
tcps_rexmttimeo | 0 | 1 | *retransmit timeouts
tcps_rexmttimeo_syn | 27535420 | 0 | *retransmit SYN timeouts
tcps_rcvpackafterwin | 0 | 2 | *packets with data after window
- | --- | --- |
UDP | --- | --- |
- | --- | --- |
- | --- | --- |
Flow Table | --- | --- |
- | --- | --- |
err_cwf | 27 | 0 | *client pkt without flow
err_no_syn | 0 | 4400 | *server first flow packet with no SYN
err_no_tcp_udp | 0 | 1 | no tcp/udp packet
redirect_rx_ok | 0 | 1 | redirect to rx OK
err_c_nf_throttled | 30378990 | 0 | *client new flow throttled
err_flow_overflow | 960449 | 0 | *too many flows errors并发连接测试
并发连接测试需要注意很多厂商偷偷的在送测设备中更换内存，以获得更大的并发连接数， 实际上最大的并发连接数通常为实际
接入终端个数*200，同时互联网上单个流平均的持续时间为16s左右，因此新建连接数*20~32基本上就是平台实际可以用的并发连接数，例如Cisco ASR1000的ESP100虽然内存支持10M的并发连接，通常我们默认建议用户使用的也就是4M左右，具体测试和新建类似，只是调长Delay参数./t-rex-64-o -f astf/http_high_active_flows.py -m 200000 -t delay=1000000000 --astf --no-ofed-check -c 20吞吐测试
吞吐测试的脚本
from trex.astf.api import *
打流看吞吐：
./t-rex-64-o -f astf/http.py -m 200000 --astf --no-ofed-check -c 20吞吐测试结果:
-Per port stats table
ports | 0 | 1
-----------------------------------------------------------------------------------------
opackets | 19756560 | 91901021
obytes | 1721689878 | 127458559534
ipackets | 89897230 | 19761521
ibytes | 124679395555 | 1722113091
ierrors | 1806494 | 0
oerrors | 0 | 0
Tx Bw | 1.27 Gbps | 92.73 Gbps
-Global stats enabled
Cpu Utilization : 73.1 % 12.9 Gb/core
Platform_factor : 1.0
Total-Tx : 94.00 Gbps
Total-Rx : 92.12 Gbps
Total-PPS : 10.21 Mpps
Total-CPS : 111.76 Kcps
Expected-PPS : 0.00 pps
Expected-CPS : 0.00 cps
Expected-L7-BPS : 0.00 bps
Active-flows : 355026 Clients : 0 Socket-util : 0.0000 %
Open-flows : 1317126 Servers : 0 Socket : 0 Socket/Clients : -nan
Total_queue_full : 58711424
drop-rate : 0.00 bps
current time : 50.6 sec
test duration : 3549.4 sec应用流控测试
这个需要多个流混跑，如下所示：
[root@Trex v2.86]然后打流
./t-rex-64-o -f astf/sfr_full.py -m 300000 --astf --no-ofed-check -c 20看流控结果
-Per port stats table
ports | 0 | 1
-----------------------------------------------------------------------------------------
opackets | 65889293 | 132685318
obytes | 24917530681 | 142126089015
ipackets | 112246897 | 65421150
ibytes | 119222904031 | 24565262063
ierrors | 33314 | 0
oerrors | 0 | 0
Tx Bw | 11.10 Gbps | 58.37 Gbps
-Global stats enabled
Cpu Utilization : 84.2 % 8.2 Gb/core
Platform_factor : 1.0
Total-Tx : 69.48 Gbps
Total-Rx : 63.65 Gbps
Total-PPS : 10.62 Mpps
Total-CPS : 368.60 Kcps
Expected-PPS : 0.00 pps
Expected-CPS : 0.00 cps
Expected-L7-BPS : 0.00 bps
Active-flows : 755305 Clients : 0 Socket-util : 0.0000 %
Open-flows : 7592138 Servers : 0 Socket : 0 Socket/Clients : -nan
Total_queue_full : 768940
drop-rate : 0.00 bps
current time : 64.8 sec
test duration : 3535.2 sec看设备识别率
ASR1009X今日技术扶贫结束
《网络设备测试艺术(2)–应用层测试》来自互联网，仅为收藏学习，如侵权请联系删除。本文URL：https://www.bookhoes.com/4844.html