On Tue, 12 Feb 2008 02:32:58 GMT, Jeff Liebermann wrote in :
Hub? Not a switch? Sending non-broadcast wireless traffic separately to all wireless clients would make no sense. Idle clients should have little or no impact on wireless network performance. What am I missing?
See also: Q. My access point (AP) accepts and connects to only one client at a time. What could be the reason? A. One possible reason could be that the max-associations parameter is set to 1 under the service-set identifier (SSID) configuration. Use the max-associations SSID configuration mode command in order to configure the maximum number of associations supported by the radio interface (for the specified SSID). Use the no form of the command in order to reset the parameter to the default value. This default maximum is 255. ---------------------------
Indeed, and in your subsequent reply you quote the max association param. for a cisco AP (255); presumably this also reflects the size of the bridging table. Just wonder about the low-end products.
Where did you find an SNMP controlled power supply? I have SNMP controlled APC UPS backup power supplies but nothing that will directly run a wireless router. (Note: I'm a big fan and user of SNMP).
Cost Effective is a complicated calculation that involves things we haven't discussed or that aren't properly innumerated. For example, the labor to do monitoring, system administration, and updates. User administration is another time burner. If there's a billing system involved, it too adds cost. From my perspective, the actual access point is only a tiny part of the cost of installing and running a WLAN, whether coffee shop or 200 room hotel. Even the cost of installation is trivial when compared to the continuing cost of administration and maintenance.
Therefore, if you want something that's really cost effect (as in TCO), then all this junk and more has to be considered. I would value the typical better than junk router as being equal to about 1 or 2 service calls. If each AP requires my personal attention, I've lost money on the sale (or if possible, I pass the cost on to my soon to be irate customer). Therefore, I spend much more time and money making sure the system can stay up, stay alive, be remotely administered, and is stuffed full of monitoring software to verify operation and detect abuse. Somewhere under all that is the cost of the AP, which can be almost anything that supports the necessary remote admin, SNMP monitoring, and possibly IDS (intrusion detection system).
Even more basic is my personal goal in selecting hardware. It's obviously to make a profit, but a close second is to keep the phone from ringing. When the phone rings with an irate customer, I have a problem. If it's because I picked a cheap junk wireless access point, I made the wrong choice. However, it the system purrrrs along, with minimal maintenance, and few complaints, I'm happy.
The trick is to not overdo the hardware. Obviously, cheap consumer grade hardware is marginal at best. However, mission critical monsters, that require certificated wall paper to setup, is also not exactly optimum. The first "wireless switch" setup that I threw together would have been much easier if I had burned a week taking classes or at least doing some serious RTFM. The added complexity of some of the high end systems are simply not worth the added effort.
A better example is a friend that sold a wireless system on the basis that it would do seamless roaming. He's now beating his head against the wall trying to make it work. Yeah, it's better than junk hardware, but that does him no good if he can't make it work. I won't mention the wireless switch install that took me a week to decode the install and setup instructions.
Anyway, I kinda prefer middle of the road hardware. It's not because they're better in any way. It's because it's a good fit for my abilities and for my customers typical requirements.
As for the external watchdog timer, I use it only in desperation. Software watchdog timers run the risk of failing if the processor goes insane. I've seen it happen in TNC's (RF modems) where the watchdog timer seems to cause more failures than it prevents. Methinks an external timer is the only really effective way to be sure it properly reboots. Since the external timer doesn't watch any kind of keep alive pulses, it's not really a watchdog. Therefore, if the device is prone to crashing, hanging, or regular abuse, such a periodic reboot isn't going to help much.
Incidentally, I have it reboot slightly BEFORE midnight, instead of at midnight, so the logs look a bit more sane.
One place where the watchdog also made an improvement was in my mountaintop weather stations. Just about everything was hanging or crashing due to multiple causes. Some of these problems are not going to get fixed until perhaps summer. So, every night, a half rack full of boxes gets a very graceless power cycle reboot. That causes a small hole in the collected data and triggers an alarm (SNMP trap) which I ignore.
So, why do I continue to use cheap wireless routers? Because cheap routers is what DD-WRT, OpenWRT, and other open source replacement firmware runs on. If it would run on Cisco AP's, in place of IOS, I might use Cisco. The ability to transplant a configuration file from one model wireless router, to another model, simply because they are running the same firmware, is a huge time saver. If one vendor (i.e. Buffalo) temporarily is unavailable, there are dozens more that will run the same firmware. I don't even look at the stock firmware any more.
Cisco wrote that, not me. I know that wireless is a "multiport ethernet bridge" also known as an ethernet switch. Apparently the author of the FAQ did not.
Nope. You are correct (this time only).
Yeah, I saw that. I think (not sure) that this varies with AP model. The FAQ seems to hop around between the 340, 350, 1100, 1230, 1240, and other models. It's difficult to tell if this is in reference to one or all of these models. Also, reading between the lines, it appears to be a limit per SSID. Since some of these models support dual or multiple SSID's, 255 may not be the final limit.
Homebrewed it. Necessary due to well-known lockups which plagued the board I used for the AP when it is under high traffic load. But for some reason the lockups have ceased after the board has been frozen at -25 degrees F for a few weeks and then returned to service when temps rose above zero. The board's installation is pictured here:
The bulk of the units listed are wireless routers. There are a few wired routers mixed in. If you check the "WLAN" box on the filter near the top of the page, it will limit the listings to only wireless routers.
Well, the only ones I know for sure (from testing) are: WAP11 31 entries BEFW11S4 31 entries. DWL900AP+ 15 entries Unfortunately, that also varies by hardware mutation and firmware version. For example, most of the WRT54G series of routers will bridge at least 63 MAC addresses. However, when the v5 mutation arrived, the table shrank to (my guess) about 7 entries. Also, I can't be sure it was the MAC address bridging table that was limiting the number of connections, or if it was some other table.
Unix boxes have virtual memory. They don't need to keep such things in RAM. When I see a commodity router with a built in hard disk, methinks maybe it could handle more MAC addresses.
Oops. I switched from an access point (bridge) to a wireless router (wireless bridge plus an ethernet router). Sorry. I meant it in reference to yet another table that can be flooded and overflow.
None. It's in the router section where the IP socket buffers are located. On the typical wireless router, there is no shared memory between the wireless chips and the router chips. That AP builds the MAC bridging table, while the router does everything else.
Exactly. An access point is a wireless (ethernet) bridge. There are no layer 3 features in a wired or wireless bridge. Well, there is the usual management interface that requires an IP address, but that doesn't have anything to do with the bridge functionality.
Yep. That's the management interface (web, telnet, or SNMP).
Wrong. The basic wireless security is handled completely by the AP. WPA/WPA2 encryption, with the various methods of authentication are all Layer 2 based protocols.
Filtering by MAC address is sometimes handled by an access point. Same with a list of acceptable MAC addresses allowed to connect. I'm not sure it's very good security, but it's there.
What do you mean by "standard hardware"? I've built wired and wireless routers based on a PC platform. I have several running on PC104 cards with CF card for a disk. There are plenty of open source PC based firewall and router projects available. There are also boards specifically designed for such platforms. Random sample:
Wireless is just another port on the firewall/router.
Good assumption as long as you don't include the other limitations. The big ones are cost and memory. I recently bought a wireless router for $35 retail. There's not much room in that price for the table space and RAM necessary to provide a large number of wireless connections. Slow processors also restrict the aggregate thruput and degree of packet inspection, that can be performed. The worst part is that the number of simultaneous wireless connections is just something that isn't sold as a feature. Few home users need more than a small number of connections. 8 would be more than adequate for most users. So, given the choice of adding yet another acronym, buzzword, feature, function, or gizmo to the wireless router, versus increasing the number of connections, the feature bloat always wins. That eats RAM and the rest you can predict. Frankly, I consider cramming a full featured wireless router into 2MB of FLASH and 8MB of RAM, to be rather impressive.
On Mon, 11 Feb 2008 21:45:42 -0800, Jeff Liebermann wrote in :
Once is better than never. :)
I think 255 is just the default value of the maximum allowed associations control, which can be adjusted higher (up to the absolute maximum of 2048) or lower than the default value.
Maybe. I did some Googling for max-associations and found nothing on a maximum limit. However, many of the examples for various Cisco access points had fairly low numbers, such as 15 thru 50 as examples. The highest I found was 254.
Cabling-Design.com Forums website is not affiliated with any of the manufacturers or service providers discussed here.
All logos and trade names are the property of their respective owners.