My Photo

My Online Status

April 19, 2008

What a Mesh!

GuyentangledinwiresWireless meshing is nothing new. 

A wireless LAN mesh is made up of a group of cooperating APs, only some of which are directly attached to an Ethernet - the rest are merely plugged into a power source.

The APs form a wireless topology to route client traffic between any member of the mesh and the wired network. Ideally, a mesh WLAN is self-forming, self-optimizing and self healing, much like a router network; but unlike a router network, mesh WLANs must cope with two uniquely challenging variables - interference and media sharing - that have largely contributed to its uncertain viability in the enterprise.

While the concept is quite appealing for many environments like hotels, schools, warehouses, airports and public access areas, poor performance, complexity and reliability have made meshing a non-starter in the enterprise. 

A packet consumes bandwidth at every hop on a mesh path. Because Wi-Fi is a shared medium, it creates a delay for other clients and backhaul connections contending for bandwidth in the same frequency channel. Multiplied by the number of hops the packet must traverse within the contention domain, this depletion of capacity greatly limits the scale of the mesh WLAN.

Regardless of meshing, enterprises have been frustrated with the limited range and unpredictable performance of WLANs. Meshing only compounds the problem - more APs are involved in completing a client transmission, more traffic competes for wireless bandwidth and there is higher exposure to the possibility of interference. 

Meshchart_2 A new approach, dubbed SmartMesh, looks to overcome this crap. SmartMesh combined directional beam steering and forming with 802.11n to provide 300Mbps of access and backhaul capacity.  Backwards compatible with existing Wi-Fi clients, a 802.11n SmartMesh reduces the packet delay per mesh hop by as much as FIVE times over an equivalent 802.11g/a mesh, and a three fold increase in UDP measured throughput at two hops away from the root AP.

Moreover, the directionality of the embedded Smart antenna array gives SmartMesh APs 50% more range (at a given throughput) over conventional 802.11n APs so fewer of them are required to cover a given area.   This in turn reduces hop count and backhaul traffic load, further enhancing systems performance.

Confronted with interference, conventional APs either drop packets or respond by lowering the transmit data rate which reduces system throughput.  Smart Wi-Fi selects a signal path that avoids the interference, thereby preventing packet loss and preserving the higher transmit rate. Should it fail to find a quality signal path, auto-topology software in the downstream SmartMesh APs will reroute their wireless backhauls automatically when they detect a drastic decrease in the potential performance of the AP in trouble.

Detection algorithms that rely on SNR (Signal to Noise Ratio) statistics reported by the Wi-Fi chipset are often flawed because most chipsets cannot distinguish nearby interference, a simultaneous transmission by a client at close range to another AP, for example, from a strong signal.  Even when a mesh AP senses the interference, its response, such as re-routing or changing of the power and channel settings, typically requires actions on multiple nodes.  By the time the network converges, the interference may have transpired.   In the worst case, this can lead to “flapping” which de-stabilizes the mesh. 

Ultimately, by combining 802.11 with intelligent RF controls, Wi-Fi meshing can fundamentally change the traditional economics of wireless LAN deployment - reducing the cost of a WLAN by 50 percent, slashing deployment time by 50 percent while delivering three times the performance of a conventional 802.11g WLAN.

Now that's interesting!

March 25, 2008

Let the N Testing Begin...

Crn_article_mar081_3 Computer Reseller News (CRN) just published some of the first competitive 802.11n testing. 

Not surprisingly, they had problems getting anyone (read Aruba, Cisco, Trapeze, etc.)  to even participate - go figure (CSCO even blogged about why they didn't show up...hmmmm).

Well, we're risk takers.  We believe you just need to build good products and put them out there. Like any vendor, we'd love to be able to control reviewers, dictate test plans, influence test results. But we're also realists. We just don't have the clout.

And 802.11n is hard to test. There are so many variables.  Meanwhile vendors do nothing but spew data rates that users will never see.  Then they never seem to show up when the curtain rises.

With 802.11g the maximum physical rate is 54Mbps, but actual UDP performance is capped at roughly 35Mbps with maximum TCP throughput still lower. When using higher 802.11n, physical data rates the problem is even more dramatic. But at a 300Mbps phy-rate with no channel bonding or block acknowledgment, the effective maximum UDP throughput for 802.11n is no more than 68Mbps.

At CRN we received top marks for performance at 20MHz with the AP doing about 66Mbps. With 40MHz channelization (in the 2.4GHz band no less) we saw TCP data rates exceeding 125Mbps - but this wasn't even reported.What a lot of people don't understand is that to see any kind of substantial performance increase, three things need to happen:

  1. spatial multiplexing/multipath (with more than one stream)
  2. 40MHz channelization (i.e. channel bonding) and
  3. frame aggregation and block acknowledgment.

But today's 802.11n systems do next to nothing to increase the likelihood that these fundamental techniques are actually utilized. This is because they have little or no control over the RF domain and the path that Wi-Fi signals take. So they can do almost nothing to mitigate interference and packet lose which are essential to ensuring spatial multiplexing and channel bonding (for instance).

Pretty much every 802.11n AP today uses multiple omni-directional antennas for different streams. These omni antennas are equally polarized, thereby reducing the likelihood of sufficient multipath, especially over short distances.

Smart antenna technology is ideally suited for spatial multiplexing. Because of the control over signal path direction and timing, smart antenna arrays maximize pattern de-correlation. It is the de-correlation of the different antenna configurations that provides the bulk of the statistical gain delivered through the use of such systems.

By independently steering or routing each spatial stream over the most optimum RF path, smart antennas exploit the utility of spatial multiplexing by increasing the percent of time that spatial multiplexing communication is possible.

Hopefully as more 802.11n tests come so will the vendors.  Like Woody Allen says "Ninety percent of life is just showing up."

March 09, 2008

Demo Diva Raises Ruckus at CableLabs...

Selinaagainstwall2With her back up against the wall at the recent CableLabs Winter Conference in Colorado Springs, our CEO, Selina Lo, was all business (sometimes she's good and sometimes she's REALLY good. This day she was in some sort of zone).

Cablelabs_2 Out of nine companies innovating new  technology for services provider, Ruckus was asked to present within CableLabs' "innovators showcase." 

So we demonstrated our latest 802.11n Smart Wi-Fi system to an audience of around 200 cable executives.  They were impressed enough to vote it the "best product idea most likely to succeed." Go figure. I thought cable companies loathed Wi-Fi?  Not any more.

For the demo (we have some of the coolest demos), we streamed three MPEG-2 high defnition multicast streams concurrently. We then made a UMA/VoIP over Wi-Fi call using the newest Blackberry Curve and T-Mobile's UMA service over the same network. 

Going for broke, we pulled up some content off a Sling Box in Lake Tahoe and surfed the Internet - all over the same 802.11n network.  See for yourself.

March 01, 2008

802.11n without Channel Bonding is Just Stupid

Choices_2802.11n is a good choice. 802.11n without channel bonding is not.

Perhaps the most important thing (in addition to MIMO and frame aggregation) that makes 802.11n 802.11n is 40Mhz channelization, aka channel bonding.

Channel bonding boosts bandwidth by combining two adjacent 20 MHz channels into a single 40 MHz channel. The throughput increase is actually a bit more than double since the guard band between the two bonded channels can also be removed.

Since only three non-overlapping 20 MHz channels are available in the 2.4 GHz (2.4 GHz to 2.4835 GHz) spectrum band used by the IEEE 802.11b and IEEE 802.11g WLAN communication standards, channel bonding is often thought to be difficult in noisy environments with dumb Wi-Fi systems that have no proper adaptive controls.

But here's the real deal. Without channel bonding 802.11n is handicaped. Limiting clients (e.g. Centrino and other laptops) to 20MHz-only is essentially misguided and actually hurts 802.11n as a whole. People that think that 40MHz operation causes extra interference don't actually understand RF issues.

Yes, 40Mhz operation consumes twice the bandwidth, but the packets are half as long (relative to time) so there is no net increase in interference. APs and clients should be smart enough to dynamically determine the correct mode of operation.  Smart Wi-Fi systems do just that.

Smart Wi-Fi systems continuously pay attention or listen to (my engineer says i must say "monitor") the RF environment, picking the best performing signal paths and RF transmission parameters for a given client.

This results in less interference since the link is occupied for a shorter period of time.  There you go.  Now tell someone who really cares.

February 08, 2008

The Shape of Things to Come?

Air Yes, thin is in. Just ask Apple.

But what's most amazing about Apple's Air is that it only uses Wi-Fi for network connectivity. There is no CD or DVD drive for getting applications loaded (yes there's a USB connector). They provide a special utility that lets you use the network to access another CD/DVD drive in a different machine somewhere on a network somewhere. And this is Wi-Fi thing only the beginning.

Instat recently estimated that by the end of next year mobile (Wi-Fi enabled) device shipments will reach 170 million up from 70 million in 2007.  Wow. Another interesting factoid centers around the use of Apple's iPhone. According to M:Metrics, for the first three months of use in the U.S., nearly 86% of owners use the device to go online. And the Kelsey Group reported that 44% of Americans would consider upgrading their phone if it provided better Internet access.  I just did.

Lodgian_logo_2c Meanwhile hoteliers everywhere are scrambling to overhaul their vanilla Wi-Fi networks. Wi-Fi is no longer an amenity at hotels but now a reason for people to choose one.  But it better work, and work good or guests will complain and complain loudly, probably never returning. Crappy Wi-Fi has become on of the top three complaints at hotels (the other two are bad-breath front desk clerks and no Ouzo at the lobby bar...I think). 

Lodgian, the country's largest independent hotel owner and operator with brands such as Intercontinental, Courtyard Marriott, Crowne Plaza, Raddisson and others, recently undertook and aggressive Wi-Fi project to outfit all their properties with a more industrial-grade Wi-Fi that could reach everywhere and support applications such as voice and video.

Wi-Fi is quickly becoming essential at home, in the office and everywhere else you go.  But Wi-Fi still sucks for doing anything other than surfing the net.  This is because it can't adapt to changes in the environment very well.  And while 802.11 speeds things up, it doesn't do much more.

For Wi-Fi to be the useful utility everyone wants it to be it must be as reliable as a wire, reach everywhere, perform like Big Ben, be as simple as any Apple product and support multimedia everything.

Once this happens, we'll begin seeing Wi-Fi used as a viable service delivery mechanism by carriers.  They already view Wi-Fi as a strategic part of their service strategy and are planning to use it for new managed services at home, hotspots and in the enterprise. 

January 20, 2008

802.11n: A Prison of Promises?

OathYes, 2008 is the year for 802.11n. But while every wireless and networking vendor on the planet has the same access to the same 802.11n silicon, what is going to differentiate one system from another?

802.11n systems will be differentiated based on technical advances that add value to these chipsets.

Unlike the previous 802.11 a/b/g standards, 802.11n comes with a whole host of new knobs (read controllable system parameters such as physical layer data rate, frequency of operation, radio noise rejection, etc.) that need to be automatically tuned and optimized to realize the technology's full potential.

Technical advances, such as smart antenna arrays and dynamic quality of service engines, will help to control and focus signal propagation, mitigate and nullify interference, and automatically tune and optimize all these new parameters and are essential to making 802.11n live up to the hype. 

Consider the Paris to Dakar Rally. The winner of this race is the one that is capable of handling the harsh environment and road conditions along the way - even though all participating cars must meet the same standards in terms of maximum engine size. Similarly 802.11n products that better adjust to a changing Wi-Fi environment, such as dealing with interference and dynamically managing quality of service will out perform others.

The 802.11n standard allows for a range of modulation and coding schemes chosen dynamically based on environmental conditions, delivering from 6.5 mbps to 600 mbps of bandwidth to the end user. It also provides for almost an order of magnitude more data rates from which to choose.

But unlike the 802.11 b/g standard, higher throughput is dependent on the ability of the Wi-Fi infrastructure to use multiple spatial streams and 40 MHz channels in a given environment.

Wi-Fi beam forming/steering technology better adjusts to environmental conditions and as such will contribute to increased performance relative to other 802.11n products.  Here's how:

  • It increases the likelihood that 40MHz channels can be used (instead of 20MHz channels).
    Other 802.11n products would be able to use wider channels but only when they are free of interference (and when ISN'T there interference?). Adaptive smart antenna technology mitigates undesired interference by continuously choosing the best and most optimum Wi-Fi beam pattern based on where the receiver is located. In turn it can use a wider channel in more situations, even if there is some interference in the area.  This means higher and more consistent throughput.
  • It increases the likelihood that multiple spatial streams can be used (instead of a single stream).
    All of today's 802.11n products (consumer OR enterprise) rely on the spatial diversity provided by simple omni antennas mounted next to each other in a very similar form. Products that take advantage of diverse beam forming and steering techniques (leveraging both horizontal and vertical antenna polarization) will be able to generate multiple streams have a much higher likelihood to be (decorrelated) better constructed at the receiver thereby resulting in a higher data rate.

With Smart Wi-Fi systems you effectively double the chance that higher data rates are used because of the ability to effectively guarantee the spatial multiplexing and channel bonding techniques that make 802.11n so fundamentally different. Controlling signal path selection, dynamically moving Wi-Fi signals to better paths on a per packet basis and automatically tuning software parameters that need to be optimized will deliver the best possible 802.11n performance.

December 13, 2007

What You Need to Know about 802.11n

Billipod Our co-founder and CTO, Bill Kish, has been doing the networking “podcast” circuit lately.

A couple of his most recent appearances have been to discuss 802.11n in the enterprise with Nick Lippis of the Lippis Report and Keith Shaw of Network World’s Panorama

Bill is a member of IEEE’s Project 802.11 and he discusses the ins and outs of what makes 802.11n such a great technology, as well as what its limitations are (that you probably won’t hear others in the industry mention since everyone likes to promote 802.11n as the networking panacea of the decade).  He also talks about a way enterprises – SMBs and the Fortune 500 – can take advantage of 802.11n’s performance benefits without upgrading their entire networks such as using ‘n’ as a wireless mesh backbone infrastructure for 802.11g clients and access points.

The need to increase the range and performance of enterprise WLANs is on every IT administrator’s mind, because they want to provide 50 or 100 Mbps of coverage everywhere all the time so employees and visitors can easily use their laptops and dual-mode PDAs wherever they are. 802.11n addresses some of these challenges, but certainly not all.

In the Lippis podcast, Bill says 802.11n is a highly complex technology, which is both good and bad.  On the one hand, the technology’s complexity means the chips companies will stay very busy and competitive for the next several years as they innovate new features and capabilities, such as multiple spatial streams, 600Mbps performance with beam-forming, etc.  Also, unlike its a/b/g predecessors, .11n is scalable, so people can upgrade their networks in an evolutionary manner and plan ahead as new features become available.

There’s a lot of misperception about 'Ns' true capabilities, and companies are confused about what to do and when.  There are still so many legacy a/b/g systems that can’t take advantage of 'N' Bill’s advice is to implement 802.11n in the near-term as a wireless backbone to aggregate a/b/g traffic at the speeds that ‘N’ offers.  By doing this within a wireless mesh infrastructure, enterprises can extend Wi-Fi to places where cabling is prohibitive of difficult to install.

In the Network World podcast Bill dispels five critical myths about 802.11n, which he says stem from the fact that because the technology has been in development for several years and has undergone so many iterations, there’s been a lot of marketing hype and conflicting messages. 

Here are the five myths Bill addresses: 

  1. 802.11n solves all things wireless 
    802.11n, for many reasons, is highly susceptible to degrading performance at distance
  2. 802.11n will offer 100 Mbps and more everywhere all the time
    For many reasons, 802.11n’s performance degradation means these performance levels will be achieved SOME of the time, not all the time – the worst case can be 10% of this, and the worst case is what people remember
  3. Interference is not an issue with .11n because it provides antenna diversity
    802.11n solves the speed equation of Wi-Fi, but not reliability, so interference is still an issue
  4. 802.11n will eclipse 802.11g as the default client access mechanism
    Eventually, this is true – the question is, how long will it take for companies and consumers to fully adopt .11n products?  There are still a lot of legacy systems out there, and they won’t go away any time soon
  5. The raw capacity of 802.11n makes multimedia support over Wi-Fi a non-issue
    Baloney.  It's not about top line bandwidth capacity, it's about delay.

Bill says 802.11n is a great piece of technology. "It’s the product of great people who are passionate about WLANs working for many years, and it will keep people busy for several more years as the technology’s full potential is realized. We just need to be more realistic about what it IS, and what it’s NOT."

October 17, 2007

The Darker Side of 802.11n

Fingersmakingbox A reporter recently posed a very interesting question to us: "If I have an 802.11g client running talking to an 802.11n access point, will I see any improvements?"

Interestingly enough the answer was an astounding "maybe."  And she hated it.  "It's too [EXPLETIVE] complex," she complained.  "This is the dark side of 802.11n that I didn't want to know about."   Here's what we told her:

802.11n uses multiple radios to drive multiple power amplifiers for transmitting data.  Typically each of these power amps would be identical to the single power amp found in a standard 802.11g client.  If you can effectively combine the power amplifiers, you can double client's transmit power. 

This can result in 3 dB of additional link budget (read you can go a bit faster or a bit farther), that's the goal at least. Turns out, you can't simply transmit the same signal on two antennas without it interfering with itself in many directions.  If you do, unintentional, uncontrolled beamforming will occur. This is sort of like pointing a high gain directional antenna in some arbitrary direction and results in random coverage gaps.

Actually realizing this 3 dB gain doubling requires a bit of an engineering trick. The trick is a technique specified within the 802.11n spec called "cyclic delay diversity." CDD essentially tries to keep the two copies of the signal from interfering with each other by constantly varying the phase of one relative to the other. Which, skipping more gory details (text us if you want them) avoids the coverage gaps associated with unintentional beamforming.

When an 802.11n client is receiving 802.11g signals, that client will use both radios and their associated antennas to figure out how to combine multiple signals it hears to maximize the signal-to-noise ratio.  Each radio on an 802.11n client can listen for the same signal and put the best of both signals together. It's like having two flawed pictures of the same thing but when you put them together you get a perfect picture.  This is what's happening.  A technique called Maximum Ratio Combining (MRC) is implemented in typical 802.11n chipsets to do all this crap.

So what's the net-net of all this?  802.11n is complex.  You have at least twice of almost everything and lots can go wrong. But when you get it right you'll see better performance and range - even when speaking 802.11g.

Now you know why she was pissed.

September 16, 2007

Pee-ing All Over 802.11n? Not Really.

PeeingThe market is fascinated with and salivating over 802.11n.  Any vendor saying anything about the technology seems to get an unwarranted amount of attention.  Cisco is just the latest. 

Call it peer pressure, but Cisco announced the availability of their first 802.11n enterprise AP - the dual-radio Aironet 1250 - touting it as "the first enterprise-focused vendor-branded access point in the Wi-Fi Alliance's pre-802.11n Draft 2.0 certification test bed." And all for the rock bottom low price of $1,299.  You've got to give it to Cisco for being as aggressive as they are for being so rotund.

While 802.11n represents a huge increase in potential top line capacity, the ugly truth is that 802.11n, as great as it is, (and we actually think it's great) won't become mainstream in the enterprise anytime soon - especially the way it's being proposed by vendors to give users higher speed Wi-Fi access.

First, it is a fundamental architectural departure from traditional Wi-Fi technology, taking advantage of new and very complex spatial multiplexing techniques. But what's more, it's too expensive.

802.11n requires new silicon that is currently 3 to 4X moreHurlston expensive than existing 802.11a/b/g silicon.  And this isn't a single chipset. It's multiple chips. 

It also mandates new hardware on both the transmit and receive ends of the connection. This means that every computer will need a new network interface card or Wi-Fi chip that supports 802.11n (with a hefty cost associated with it). Broadcom recently said as much when it revised its outlook on 802.11n chips.

A much more pragmatic application of 802.11n will be as a backhaul wireless technology for aggregating 802.11g client traffic through a meshed wireless network (click on Meshdiagram_4diagram to see how it will work). Now that there's technology (eg. smart antenna arrays) that can continually select the best path for Wi-Fi signals, you can effectively create dependable high-speed wires through the air to enable a reliable meshed LAN without Ethernet cables at all.

Dual-band (2.4 and 5 Ghz) 802.11n can be used to provide high-speed connections between APs and 802.11g connections to clients. But initially 802.11n nodes can be used as root APs (APs wired to the LAN but communicating wirelessly to other 802.11g nodes) while still providing some 802.11n client access as needed (for those that really want/need it).This will keep costs down (read fewer 802.11n APs) while providing a simple, easy, smooth (insert your own adjective here) migration path to 802.11n.

Just like gigabit Ethernet was used to aggregate fast Ethernet connections, 802.11n will be used to plot out the same course in the wireless world. It's a much more practical implementation of the technology that leverages the big pipe that 802.11n provides while justifying price tag it demands. 

Ruckus, the voice of reason?  You heard it here first.

September 03, 2007

Corporate America: Get a Life!

Dancing Companies need to get a clue-train (see book below). 

They must begin to realize that they are effectively a person with a personality. They have a voice. And they live and they breathe. While companies are made up of lots of people, that collective is a singular entity in the mind of, well, everyone.

Most people like to be with other people they like (or people who are like them).  The people who buy networking gear are no different. They're just people.

CluetrainThis same concept is true with businesses - it's just that businesses don't get it.  Most people want to do business with companies they like, companies who treat them right, are responsive, fun and interesting. See these 95 cluetrain themes.

When customers complain, most companies get all bent out of shape and think that the customer is a whiner or jerk.  It's just the opposite.  Customers complain (most of the time) because THEY CARE. They want the products to be better. They feel invested - invested enough to take the time to say something. It might not be what the company wants to hear but at least they care enough to say anything.  It's when customers don't complain that companies should get worried.

Given the choice between two identical products, people will choose the companies they want to do business with. In fact, customers will often select an inferior product if they like the company. The vast majority of companies don't really understand this.  They are cold, corporate, boring, stick to convention and unimaginably predictable.  They think you're either FOR them or AGAINST them - friend or foe. There's no middle ground. But life isn't like that.  And we've taken this to heart.

It's the little things that make a big difference - like providing a CHAT button on your Web site to get to a live person (even if that person can't immediately help you), it's overnighting product to someone who's anxious about getting something resolved (even if that anxiety isn't shared or justified by the company), it's admitting that your products aren't always as plug and play as you say, it's making cell phone numbers available so someone knows they can talk to someone, anyone (when it's convenient for the customer NOT the vendor), if there's some question or issue. It's letting people try out your product because they're scared it won't work right.  Stuff like that.  Customers want to deal with people, not big companies.

Companies need to begin thinking like they are a person - and decide what kind or person they want to be (who would you rather do business with: George Clooney or George Bush?)....exactly!  Now you see.

When some issue comes up, instead of getting a 20 people in a room to perform mental masturbation on some problem at hand, companies merely need to step back and say "if this company was a person, what's really the right thing to do?" Then do that.  Sounds simple, it isn't.

There's basically two kinds of people (and hence two kinds of companies) those who look at a problem or opportunity and can give a million reasons why it can't be solved or it won't work and those who immediate say "YEAH!" then give you a million possibilities of how to get something done.

We're the latter (at least we like to thing so).  Hell, we named our company Ruckus Wireless not Ingenious Networks.