WaveRider Dynamic Polling MAC Loading
Transcription
WaveRider Dynamic Polling MAC Loading
WaveRider Dynamic Polling MAC Loading Introduction This brief is an example of a current WaveRider wireless system running VPN’s across its airlink. Terminology CCU3000 – base station radio or single access point EUM – subscriber wireless modem, includes antenna DPMAC – dynamic polling media access control GoS – Grade of Service Background The Media Access Control (MAC) layer determines which EUM may transmit and when it may transmit. Through the MAC layer, the CCU determines which associated EUM may transmit next and indicates to the EUM that it can transmit by polling it. The frequency with which an EUM is polled is based on its assigned Grade of Service (GOS). The CCU transmits a directed poll to the EUM, which immediately transmits a response to the CCU. After the response is received from the EUM, the CCU transmits the next poll. In this way, the inbound (EUM-to-CCU) and outbound (CCU-to-EUM) channels are maintained collision free. If the CCU has data to send to an EUM, then that data is sent with the directed poll. If the EUM has data to send to the CCU, then that data is sent with the EUM response to the poll. EUM’s that are not authorized are not polled. To optimize polling efficiency, EUM’s that no longer have traffic to send are not polled. EUM’s that are not being polled can submit a request to be polled by responding to a special random access poll transmitted regularly by the CCU. Collisions may sometimes occur on this random access channel; however, since only a small number of users are vying for service through the random access channel at any one time, the effect on channel performance is negligible. Recovery from these collisions is made possible by random back-off and retry. Once again, if the EUM requesting service through the random access channel has data to send to the CCU, it will be included with the request message. If the CCU has outstanding broadcast messages to send, they will be sent to all EUM’s with the random access poll. An automatic repeat request (ARQ) scheme, using acknowledgements and retransmissions to recover from message losses due to collisions or radio link errors, provides reliable transport. Each transmitted data payload is numbered in the packet header. Each packet header also contains an acknowledgement for the last correctly received payload, by number. If a CCU or EUM does not receive an acknowledgement for a payload that it has transmitted, it retransmits that payload with the following poll of, or response from, that EUM. A payload is transmitted a maximum of four times, after which it is discarded. Note that contrary to the 802.11system, MAC-layer acknowledgements are not transmitted as separate packets, reducing overhead by 33%, on average. Wireless Network Example – Commercial System Page 1 of 9 This system is operated by a WiSP with seven CCU3000’s in an area of about 10 square miles. It is in a town with a population of 10,165. The subscriber mix is one of residential, SOHO and small business. They currently have approximately 500 subscribers online. There are 51 registered subscribers on a single access point modem in this example. Of the 51 registered users, 13 are running a VPN service. Even though the IP address scheme is private, the operator uses NAT to provide public IP addressees to allow for the VPN service. The following is a table that depicts the grade of service levels with number of assigned subscribers. Service Class Label Best Effort Bronze Silver Gold Active ftp Rate (kbps) 15 - 394 9 - 1300 9 – 1460 18 - 1460 Number of Subscribers Assigned 2 24 0 25 Registration Table from CCU (access via CLI or SNMP) REGISTERED EUMs EUM ID GOS Class RSSI dBm SQ RNA dB Time s Rx-Octets Rx-Packets Tx-Octets Tx-Packets 60:05:d9 bronze -82 9 22 67 11097088 60700 63411780 60:05:ef bronze -85 10 9 23949 504926 3816 5658923 4745 60:07:f6 bronze -77 7 15 14 1411712 10276 9235339 11825 60:08:37 gold -75 4 27 0 682122218 4274995 2.816E+09 4308820 60:08:e4 bronze -79 17 27 30145 5003969 19115 12848654 18738 60:09:39 gold -74 6 29 134 7691192 25417 26748755 28996 60:09:4c gold -86 21 19 59 18030439 134256 178145864 173261 60:09:81 bronze -85 9 22 42 4337659 26937 18327021 28225 60:0a:23 gold -88 11 14 623 754139 12328 2052359 17819 60:0a:37 gold -84 9 21 0 38089746 288503 152281594 284675 60:0b:52 bronze -82 7 19 7312 5796968 39420 54373564 52076 60:0b:a1 bronze -85 6 20 1883 16791188 109709 114112886 138766 60:0d:a4 gold -85 7 21 28321 1148088 7445 4432872 7502 60:0e:05 bronze -75 9 30 0 28188255 62325 17952197 60185 60:0e:34 gold -85 7 22 20769 216572 1028 900922 1180 69192 60:0f:bb gold -85 7 22 1 44944010 228757 189667518 239250 60:12:12 bronze -80 7 25 28 1312727 9319 10014769 10897 60:12:4a bronze -85 8 22 2 54800 547 70387 535 60:12:59 bronze -84 7 22 12223 4629153 33329 37661734 41539 60:13:05 bronze -86 6 7 8202 706134 6959 12676525 10035 60:13:16 gold -85 4 21 761 623150 8125 13701507 11829 60:15:3c gold -85 9 21 25 8524168 57432 45830596 58091 60:15:99 gold -84 8 22 688 5811789 45384 66802783 59531 60:18:50 gold -78 16 28 11 56797333 366734 522116239 481368 60:18:6f gold -84 11 21 4972 214575639 949650 316932699 837441 60:18:70 gold -84 12 22 1356 573237 4631 1075082 4569 60:18:96 gold -80 10 25 3 407844600 3369046 3.474E+09 4065926 60:18:dc bronze -85 10 21 8823 76738 105 34097 85 60:18:ef gold -85 21 21 4 1864438 14826 16518864 18458 60:19:5e bronze -86 5 21 181 4500879 29168 29692847 33070 Page 2 of 9 60:25:0e gold -81 7 25 2 957424 14564 64329 350 60:25:7d bronze -76 8 31 3521 8600649 63448 42830548 65033 60:27:af gold -79 6 27 144 15272422 86618 73919147 99400 60:28:75 bronze -79 11 26 16077 358266 2495 2380617 2762 60:2a:45 bronze -78 6 28 5313 2510854 16866 19742870 22339 60:2a:87 gold -77 7 29 11 187173958 1523630 2.178E+09 1978023 60:2a:8f gold -81 6 25 5438 4946980 20461 12951304 20991 60:8a:d8 bronze -87 7 18 387 199160 2122 5178752 3553 60:8b:0b gold -83 8 22 224 4663547 9189 5523869 9141 60:8e:61 gold -84 7 22 0 53396617 124353 10086007 83368 60:95:60 bronze -77 5 28 4267 113612 396 259788 394 60:95:e0 bronze -84 6 22 2742 892846 2579 2101421 2614 60:96:37 bronze -79 4 12 995 2114609 13353 11767281 14067 60:a4:5f gold -78 6 27 22 1822271 11419 16402399 14173 60:ab:73 bronze -76 7 29 20333 77203 725 1176594 999 60:af:a4 gold -83 5 23 5 10396408 67860 68844587 77447 60:af:ba gold -79 4 26 1 2975915 29124 20387662 29567 60:b0:7d bronze -84 5 22 5967 1428986 13394 19301251 17895 60:b0:ed bronze -80 5 25 17472 610113 3660 6662504 5172 60:c8:e9 be -86 8 20 20597 617152 3649 5134762 4632 60:c9:53 be -88 8 14 3283 71040397 250109 29787543 254736 Monitoring Channel Loading Using WaveRider’s radiometer, the following charts show the dynamic capability of the polling MAC in real life use. 1) Instantaneous polling rate: the polling rate is steady at the maximum allowed throughout the display. Page 3 of 9 Page 4 of 9 2) Active traffic shows that the system is seldom loaded above 1/3 capacity and averages less than 1/10. The most interesting thing to note is the #EUM’s total doesn't go above 13 and averages 3 to 4. Page 5 of 9 3) Activations shows EUM’s getting to active state - this is how dynamic the system is, 2 to 4 EUM’s becoming active every second. Page 6 of 9 4) % Empty polls indicates that users are not being limited by the GOS cap - lots of small, transaction type traffic going on. Page 7 of 9 :4 0 15 :00 :4 0 15 :07 :4 0 15 :13 :4 0 15 :19 :4 0 15 :25 :4 0 15 :31 :4 0 15 :37 :4 0 15 :43 :4 0 15 :49 :4 0 15 :55 :4 1 15 :01 :4 1 15 :07 :4 1 15 :13 :4 1 15 :19 :4 1 15 :25 :4 1 15 :31 :4 1 15 :37 :4 1 15 :43 :4 1 15 :49 :4 1 15 :55 :4 2 15 :01 :4 2 15 :07 :4 2 15 :14 :4 2 15 :21 :4 2 15 :27 :4 2 15 :33 :4 2 15 :39 :4 2 15 :46 :4 2 15 :52 :4 2: 58 15 16:04:35 16:03:40 16:02:46 16:01:52 16:00:58 16:00:03 15:59:09 15:58:15 15:57:20 15:56:26 15:55:32 15:54:38 15:53:43 15:52:47 15:51:53 15:50:59 15:50:04 15:49:10 15:48:16 15:47:21 15:46:27 15:45:32 15:44:38 15:43:44 15:42:50 15:41:53 15:40:59 15:40:04 15:39:10 15:38:16 15:37:21 15:36:27 15:35:33 15:34:39 15:33:45 5) Channel utilization, the first chart shows 30 minutes of monitoring; the second chart shows over 3 minutes. 1 0.9 0.8 0.7 0.6 0.5 GO Util BR Util BE Util BC Util 0.4 0.3 0.2 0.1 0 0.5 0.45 0.4 0.35 0.3 0.25 0.2 GO Util SI Util BR Util BE Util BC Util 0.15 0.1 0.05 0 Page 8 of 9 Conclusion The system is very lightly loaded with 51 users. The number sharing the bandwidth averages 3 to 4. Page 9 of 9