Hello all,
I am new to this forum, but have been deploying Canopy since it was first released.
I would like some input from the group re: setting up a test to see what the capacity is in regards to VoIP over the Advantage platform.
Motorola has a white paper on the subject done last year with an independent research group, but the test that they set up was too sterile. It did not reflect real world scenerio. Also it was not done with the advantage platform.
As VoIP becomes more popular, we as providers need to know what the limits are to the system. We need this to establish a business model. There will be VoIP gateway providers that will want to offer their service over our Canopy networks, but due to the poor showing of other Wisps using other platforms, quality of service will be an issue that will continually come up. If we can bundle data and voice into an offering, and do it for less $$ than the customer is already paying, we have a potential home run. We will also reduce customer churn because when the customer chooses to leave, he will have to replace 2 services. But, the voice service has to be a high quality offering if we are going to be a serious player in the game.
I have quite a few customers who already use vonnage, and callvantage and all seem to work OK.
What I don’t know is what will this increase use of VoIP do to the overall network. How will it impact data throughput.
I want to set up a test using Advantage APs and regular SMs. The goal is to see how many voicec alls can be made over the network through the same AP, and what does it do to bandwidth. I am aware of the priority channel, and will set it to 10% for the test. Currently, it is set to zero, and customers have not expressed any concerns. If the Advantage AP is a 20Mb radio with about 15Mb thruput, and the priority channel is set to take up to 10% (1.5Mb) then at 20Kb per call we should see the AP handle up to 75 simultaneous calls. If good quality of service is oversold 4X, (not everyone makes a call at the same time) then the AP would have a capacity of 300 VoIP users. The limits as I was told was that each AP can handle up to 200 users.
Is there anyone out there that has pushed the Advantage platform to the breaking point?? If so, where is it?
For our test, we will be setting up 3 SM’s all going to the same AP which currently has about 70 SMs already associated and pushing traffic. We will set it up at the VoIP gateway location, and have up to 5 calls and data transfer off of a single SM, and the other 2 will be single calls and data transfer as well. We will be monitoring call quality as well as ping times, and transfer rates. We will also look at MRTG graphs of the AP and BH.
Any suggestions as to test parameters would be greatly appreciated.
Some thoughts:
Use caution with the high priority channel, if you use it at all.
I think you are going to love what the Advantage AP, hardware scheduling, Release 7.0, and 2X operation do for you.
Please do tell how your tests go.
Some background:
With hardware scheduling, the high priority channel is enabled per-SM, and is dynamically sized (you don’t set a percentage). If enough packets have their minimal-delay bit (TOS bit 3) set to 1 by some external equipment attached to the SM, they will use all the bandwidth up to the MIR values you have set for that SM. And if the external equipment is thinking in terms of differentiated services rather than TOS, it may or may not set TOS bit 3 to 1 when it sends what it considers “high priority” packets.
With software scheduling, the high priority channel is a enabled per AP by setting a fixed, uplink percent.
With hardware scheduling, the high priority channel uses a virtual channel, just like the “low” priority channel does. Every SM enabled for high-priority reduces by one the number of SMs that can be supported by the AP. In the limit, that means an AP can host ~100 SMs enabled for high priority each using 2 virtual channels, thereby consuming the ~200 virtual channels available.
P8 series SMs do not support a high-priority channel when running hardware scheduler. P9 series SMs do.
With software scheduling, the number of service requests can be a limiting factor. 3 control slots (the default when high priority % is set to 0 %) could be too little if lots of SMs are asking for VoIP packets to be transferred to the AP. The number of control slots can be increased, but then that takes away from the total bandwidth ever available.
With hardware scheduling, any slots not being used for data in a given frame are fair game for the SMs to use to make service requests. You don’t set a fixed number of control slots, but rather “reserve” a minimum number (down to 0) that then can’t be used for data and are always availabe for SMs to contend for service requests. Even if you reserve 0 control slots, there is a good chance that some frames will be stuffed with data followed by several that have many slots that are fair game for service requests.
Some VoIP gear has a setting that ensures VoIP packets get sent (to the SM) before other traffic (web surfing, downloading, etc.). Setting that plus giving each VoIP user a small amount of uplink CIR may help ensure VoIP packets get through with low latency even when things get busy. Something to at least consider - I have seen no field experience with this yet.
The 14 Mbps of aggregate throughput at an AP is only accessible to SMs running at 2X (“close-in” series P9 SMs). All of your series P8 SMs, plus any P9 SMs further out, share 7 Mbps of aggregate throughput. (This bodes well for the future as more P9s get deployed.)
Web-surfing and file-downloading are heavily asymetric activities, probably even more than 75% downlink. If symmetric traffic (VoIP) becomes a bigger part of the traffic mix, the downlink % may need adjustment.
I have been doing a lot of reading in the new manual, and this forum to get a grip on this issue. The VoIP provider that I am working with is concerned that if they form a strategic partnership with us, will they overwhelm our access points with VoIP customers (who also will be transfering data, web surfing, file downloads,etc.). This could cause poor performance on our Canopy network, and make their product look bad as well. I am contemplating becoming a reselling agent of their service over my network, and making a few bucks along the way. Better us ISP’s get a piece of it, then send it all to ATT or Vonnage.
The concern is how Canopy handles the smaller VoIP packets, and at what point will the Advantage AP’s be overwhelmed, hence the capacity of each AP will be defined. Is the capacity issue a function of bandwidth, or time slots, or frame and packet size? Or a combination of all? Where do I look in the stats to see what is going on? Bottom line, can a business case be built to become a strong VoIP provider over Canopy network, and maintain profitability. If 50 VoIP calls overwhelms an AP, then we can’t make any money, and our network integrity will be compromised, therefore I can’t offer the service. Obviously, I can’t stop someone from using Vonnage or some other provider, but I can’t offer it myself (and make a few bucks as well) and support the product if it doesn’t work.
Some preliminary testing showed that each VoIP call (ATT Callvantage) using a Dlink and Linksys VoIP routers both going through the same SM making 2 simultaneous calls showed each call to take about 85Kb. When an FTP file transfer and HTML file downloading were introduced, it brought the SM down, and it was not acceptable. BAM was set to 640 downlink and 256 uplink. Seems the FTP transfer was the system killer.
My APs are on a mountain top overlooking the Rio grande valley, and I am in a high RF area. Most of my customers are beyond 2 miles, some as far as 12+ miles, so 2X will not work for hardley any of my customers. That leaves me with software scheduling. AP’s are all advantage and everyone is upgraded to 7.0 firmware. SMs are all P8
I am contemplating setting the priority channel to 10%, and setting downlink on the APs to 65%.
Any advice is greatly appreciated.
Thanks