Let me start by saying
point-blank. Don’t stress about microwave latency. It’s not an issue, period. It’s good to ask about it, but don’t get talked into buying one radio over another because of some (supposed) latency advantage. Hear
me out and tell me if I’m wrong…
My first brush with microwave latency was 1986, so I'm pretty familiar with the topic. My engineers were writing the industry’s
first wireless Ethernet specs (typed up on my 128k MAC) and calculated
round-trip propagation delay through the air, along with whatever delays were
associated with the equipment. Those specs were confirmed by Cisco when we co-developed the first full-duplex Ethernet interface, and later by Motorola in a technology transfer deal.
Here’s what I know. Latency of a
microwave transmission is (considerably) less than optical fiber, but only as the signal travels through air as opposed to glass.
There’s also the fact of equipment latency, and that's where fiber wins. Yet microwave hardware (of the FCC licensed variety) introduces mere microseconds of delay, a minuscule figure when you consider that it takes 150 milliseconds before VoIP suffers.
So why all the fuss about latency?
Well, not long ago microwave vendors woke up to a crazy niche application for high frequency trading (HFT) and there’s been plenty of press about how every microsecond means insane amounts of money when those trades are moving between stock exchanges. I repeat, “when those trades are moving between stock exchanges.” That's calculating distances over hundreds of miles, not to a data center across town or to your remote office. The other day I noticed vendor literature that claimed to meet the needs of the "Low-Latency Market". But where is that market? What else but high frequency trading needs to obsess about shaving a microsecond here or there?
Well, not long ago microwave vendors woke up to a crazy niche application for high frequency trading (HFT) and there’s been plenty of press about how every microsecond means insane amounts of money when those trades are moving between stock exchanges. I repeat, “when those trades are moving between stock exchanges.” That's calculating distances over hundreds of miles, not to a data center across town or to your remote office. The other day I noticed vendor literature that claimed to meet the needs of the "Low-Latency Market". But where is that market? What else but high frequency trading needs to obsess about shaving a microsecond here or there?
If microwave traffic
between stock exchanges constitutes a “market”, then there’s got to be a
similar sized one that wants zebra striped antennas. Consider that there are only so many
microwave paths you can carve out between any two destinations before you run out of
frequencies or tower space and start veering off course (adding latency). Enough
about this.
I get that everyone has to
reinvent how they sell things after old pitches get stale. But let’s get real
and not try to tell enterprise users that they should buy one radio over
another because of a latency difference that offers no real advantage. Microwave is already an
inherently low/ultra low latency medium.
Believe me, when you’re buying microwave there are more important
product distinctions.
Here’s another Fact. Few sales or marketing guys – even the most technical, actually know the latency numbers of their
radios. If you really wanted to know that, you’d have to break out every radio
feature you’re getting (or enabling), and ascribe to it the amount of latency
that it introduces. Adaptive modulation? That might
cost you 50 microseconds. QOS? Maybe
20 mics. Except for high frequency
trading, I’m not about to revert to primitive radio designs to shave microseconds here or there.
Truth told, if you count hardware
at both ends, a fiber connection is still going to have lower latency than a
typical (not for HFT) microwave link because fiber interfaces can get down into
the nanoseconds and that makes up for microwave’s 2.5 microsecond/mile
advantage.
On a final note, there’s one
distinction to be aware of. Wireless under the category of "FCC licensed microwave" has far lower latency than 802.11, license-exempt wireless that runs for instance, in the 2.4 and 5GHz bands. The reason
is simple. Licensed microwave is built on solid state hardware, whereas license-exempt radios may be driven
entirely by software, and that’s where you get latency in the 30-50+ millisecond range. 150 milliseconds (one-way) and VoIP starts to crap out.
There's nothing on the planet that's lower in latency than old-school, analog (licensed) microwave. Hmmm... If I already didn't have enough bad ideas, I'd dust off my old radio designs.
#microwavelatency #wirelesslatency #wirelesswanlatency #voiplatency #pointtopointmicrowave
#microwavelatency #wirelesslatency #wirelesswanlatency #voiplatency #pointtopointmicrowave
This has the capability of speeding up the process of data transferring and thus fitting the data in a packet size.
ReplyDeleteUltra-low latency MAC
Thanks
Your blog provided us with valuable information to work with. Each & every tips of your post are awesome. Thanks a lot for sharing. Keep blogging,
ReplyDeleteGE JVM7195SFSS
Thank you because you have been willing to share information with us. we will always appreciate all you have done here because I know you are very concerned with our. microwave convection oven combo reviews
ReplyDelete