Slashdot Mirror
Concerns Over Increased 802.11n Power Usage
alphadogg writes "Next-generation 802.11n systems promise to considerably improve WLAN performance. But the processing required for the boost sucks up more power than the older 802.11a/b/g networks. Still, many enterprise-class Wi-Fi vendors claim to deliver full 802.11n capabilities without enterprise customers having to touch their power infrastructures. So what gives?"
I am currently reading slashdot and typing this comment on a laptop and a 802.11n wifi connection.
I just skimmed the article, but the power issues seemed to be at the switch. Are they saying that the remote clients aren't affected? I'd be more concerned about sucking even MORE power out of laptops and PDAs than out of an outlet.
I swear to God...I swear to God! That is NOT how you treat your human!
Enterprise hardware does not use general-purpose CPUs, it uses special-purpose ASICs. These are lower power than general purpose hardware. They are fab'd using a newer process than the older ones, and so use less power per transistor than the old chips. Less power per transistor means more transistors (which means more processing power) per watt. If you rolled out 802.11g infrastructure four years ago, you can now fit around three to four times as many transistors on the same area of silicon as was possible when you deployed your current infrastructure.
I am TheRaven on Soylent News
802.11n causes cancer.
It's like 8 watts instead of 3 watts (not exact numbers). It's not a significant amount of power. That's why you don't need to upgrade your infrastructure.
I'm not sure how a silly article like this gets published. If it was tons of power, how could they make 802.11n adapters for laptops?
My first thought was not for the infrastructure (the Access Points in a hundred conference rooms and spaced over all the cubicle farms). It was for the laptops. Not having one of the newest sexy 11n devices in my laptop, I wondered if MacOSX or Windows managed to drop the speed for battery operation vs tethered DC power operation.
But the article IS about the corporate fixed infrastructure, right? Are we talking 5% increase of power for something that is already only 1% of the facility power costs? Wow, going up to 1.05% of the facility power costs is gonna blow the budget and turn Antarctica into a tropic paradise. I bet most of the energy is in the form of heat from the individual DC converters plugged in to long extension cords laid along the drop ceiling to the nearest electric column.
[
First of all, this has little to do with what is usually considered power infrastructure. This has to do with power-over-ethernet. It appears some dual band 802.11n radios require more power than one particular specification (802.3af) allows. Solution? Don't use 802.3af, or, don't use the radios which require too much power. Not really a big deal. I expect that manufacturers will bring the power requirements down to allowable levels over time.
Essentially just because vendor A has devices which consume "up to 18 watts" it doesn't mean that nobody can build devices which take less than 13 Watts.
That's just 5 watts difference. You could probably achieve this by switching to higher efficiency components. Or you could store some energy for the short bursts of transmission, getting a steady power of 13 watts.
Keep in mind that most vendors probably still have the very first itteration of hardware. It will significantly improve over the next years anyhow.
It's not ironic! It's creepy. (God, kids these days.)
Are you adequate?
Multiple radios in simultaneous operation (Read: MIMO) plus the circuits to coordinate and control the radios and it has *double the bandwidth* (in the physical layer) = more power use by the 802.11n device. Duh! (Also notice there are more and larger antennas?) http://en.wikipedia.org/wiki/IEEE_802.11n
If the big deal in the article is over PoE powered 802.11n solutions, just exceed the power spec at the power injector and use 24 Gauge CAT6 UTP (or larger Gauge CAT6 for longer runs) for your PoE runs to lower electrical resistance.
I have installed PoE devices that have their own proprietary power injectors that exceed the PoE power standard. The problem is where people use long runs of super cheap CAT5 and lots of punch-downs and they expect also their large switch with PoE injection to provide PoE to whatever is connected. There is a reason that the manufacturers' of powered by PoE devices do provide their own wall-wart PoE injectors...
Unless you're running at near maximum electrical load, a few extra watts is not going to set transformers afire and melt power lines. If you were in Cuba, with its infamously collapsing electrical grid, I could see this happening, but there are no enterprises in Cuba to begin with, so it's not a relevant scenario.
It also might help that most enterprises' employees are at their desks, not wandering around on the WiFi network.
Slashdot: Playing Favorites Since 1997
Enterprise hardware does not use general-purpose CPUs, it uses special-purpose ASICs
Actually they don't - take a look in a high-end AP some time. "Enterprise" wireless systems use the same, or often older generation, of wireless technology that is in consumer access points. Competition in the consumer AP market is what drives all the incredible price/performance in wireless technology, and I assure you nobody is going to spend the tens of millions to do a custom spin of one of those chipsets for the relatively small high-end market. Those products sell on branding, special software features, and support contracts, not silicon performance. And as far as the CPU/memory etc, these are going to be much LESS specialized in a high-end system than in a consumer AP. Low-cost APs use highly integrated ARM or MIPs-based SOCs that are designed for sub $20 BOM cost. A higher-end system, however, is not bound by BOM costs and might have four times the memory and a more general purpose processor capable of running more software.
Routers and switches are a different story, and those DO use ASICs and FPGAs. The high-end models of these have to deliver a totally different hardware feature set than consumer equipment, and unlike wireless technology, the bleeding edge tends to be developed for the highest priced products before trickling down.
Does this mean I will have to upgrade the strap on my laptop bag? Will this help me lose weight? I better join a Gym and get in shape before I upgrade. This tech stuff is really difficult.
Just because a given access point will use "up to 18watts" does not mean it will always use 18watts. As long as the average power consumption is under what the supply can deliver, all can be well. Several possibilities exist.
Best case is that the supply can deliver short term bursts of power sufficient to meet the demand. This is realistic in most scenarios today, as supplies are typically rated by long-term average power. For example, the 20amp breaker on your typical home circuit will easily supply a few seconds at 30amps. My irrigation pump is rated at 28amps, the breaker at 50amps, and the pump draws over 100amps every time it starts. No problems.
Worst case is the access point requires some local power storage to meet the demand. This might take the form of a supercap or rechargeable battery. Average power usage keeps the battery charged, and peak demands are met from the battery. Only if the battery becomes depleted does the access point have to limit power usage -- perhaps by limiting transmit power or by limiting speed.
sdb
Looks like I will need to find a ZPM to power the thing and the ZPM is where they get you.
Tsukasa: All I really want, is to be left alone...
I know Cisco's 802.11n AP's use 18 watts. This is the brand most used in the enterprise. They can get it down to 802.3af by using only the 2.4ghz radio. But when using the 2,4ghz and 5ghz radio in the AP simultaneously you need power injectors, power to the AP, or new switches.
There is several ways vendors are getting around the power issue with .11n, by only having one antenna or by doing some software trickery where instead of getting 15.4 watts over every interface you lower the power to a couple and redistribute that power to get it to 18watts , (I know Cisco is doing this on their 3750-E switches and have been told they plan on it on their 4500E switches). Since most companies don't put 30 access points on one switch anyway (would be bad if that switch went out!) that probably will not be an issue, unless you run IP phones that eat a lot of juice also.
What I see as the large issue is that a significant amount of people have only 10/100 to the copper port and to run 802.11n you must have 1Gb copper to provide full bandwidth. (then power the device by how ever you want PoE, injector, wall plug)
Actually you're wrong. Cisco makes their own ASICs for their AP's, or at least they did in late 90's early 2000's when I supported their wireless division (formerly Aironet Communications). They were making enough cards and AP's that it was deemed worth it to spin their own chips. The biggest reason is they could do crypto at full wire speed in an AP that was powered by PoE, something that their competitors couldn't do at the time. I don't know if they've gotten away from ASICS as process improvements have enabled general purpose CPU's to do more with the power budget, but I'd bet they've found a new reason to spin an ASIC, like per client crypto keys.
There are 4 boxes to use in the defense of liberty: soap, ballot, jury, ammo. Use in that order. Starting now.
Actually you're wrong. Cisco makes their own ASICs for their AP's, or at least they did in late 90's early 2000's when I supported their wireless division
You're saying I'm wrong but admitting that your information is ten years out of date. A LOT has happened since then. The whole industry has consolidated and everything is now driven by the consumer AP market. Encryption is now done in hardware even in the cheapest g/n chipsets. These is NO reason to do custom silicon to support any of the features offered by "enterprise" APs.
Actually my informations is only 5 years out of date and from a quick Google search the 1300 series AP which is their next to newest offering is also ASIC based, so I wouldn't consider it so out of date =) They considered the stuff they could build into the ASIC to be a competitive advantage and knew that they could be to market with a new product months before the major chip houses would have even engineering samples. To be more on-topic their 1250AP which is 802.11n draft 2 compliant is only supported with one radio if powered by a 802.3af switch but you can use two radios if you have a Catalyst switch with 18.5W power capability or if you use the power injector. They claim 16.9W max power draw with 2 radios or 12.95W with one radio.
There are 4 boxes to use in the defense of liberty: soap, ballot, jury, ammo. Use in that order. Starting now.
Cisco has new Premium PoE blades for the 4500 E-Series chassis that provide up to 30W of power per port. As for other comments already posted, it has nothing to do with how much power the ASIC or general purpose chips in the AP's consume, that is minimal at best, and you are not going to get your power savings there. The additional requirements are for powering the additional radios in the AP's, and pumping out that power over the air. There's basically no getting around that. Some vendors say they can, but their AP's are apparently not fully powered.
I'm not a fan of analogies, but I suppose it would be like upgrading from a VW Bug to some super-duper sports vehicle, while keeping the same engine. It may look all nice on the outside, but it is just not going to perform without the horsepower (Watts) under the hood...
Siemens just announced 802.11n which does user only 12.6W
The simple answer is, "Vendors Lie".
FPGA != ASIC I've taken a lot of cisco gear apart; very few have had real, custom logic in them. (and that was very expensive gear.) Bugs in ASICs are very difficult and expensive to fix. FPGAs are simple to reprogram; that's why they show up in everything.
Cisco ceased being relevant a LONG time ago for most people. They still have their place in the ultra-high-end optical market, but most offices are best served by a bunch of big dumb switches and PC-based routers.
-Billco, Fnarg.com
No, these were ASIC's. I know because they talked about cost to have a mask made and the minimum units expected in order to justify doing one.
There are 4 boxes to use in the defense of liberty: soap, ballot, jury, ammo. Use in that order. Starting now.