Home    Bloggers    Messages    Polls    Resources   
Tw  |  Fb  |  In  |  Rss
Tim Kellogg

Telemetry Protocols, or Why HTTP Won't Work for IoT

Tim Kellogg
Page 1 / 2   >   >>
Davidmicro
Davidmicro
6/13/2014 3:21:42 PM
User Rank
Worldwide Wizard
Re: Alternative Perspective
Let us push MQTT to the different level.

Quantities analysis could be brought for between MQTT and HTTP. Parameter could be estimated in the protocol level and for example, the time that message is arrived could be analyzed among two in the same location. This type information could be valuable to get the pro and con of each protocol.

50%
50%
Tim Kellog
Tim Kellog
6/13/2014 2:18:31 PM
User Rank
Blogger
Re: Alternative Perspective
Yeah, I understand. My main problem was that he measured MQTT against his own expectations of an enterprise message queue and found it to be unfit. The thing is, MQTT is meant to be used as an aggregator of traffic from constrained devices that are common in IoT. It was never meant to be an enterprise message queue.

The public reaction to Clemens' blog has been "Oh my gosh, MQTT isn't good for anything", which is completely and entirely false and I've had to spend countless hours dispelling these myths. It's been very successful at what it tries to do, so it's very frustrating when a well-respected architect labels it as "unfit". If the public wants MQTT to be more of an enterprise message broker then the committe can consider making appropriate changes. Honestly, I don't see much reason to make those changes since there are already so many good enterprise message brokers already available.

I do appreciate your thoughts about HTTP being the language of the web. Honestly, I think this deserves another blog post for clarifying my position on it. I share a lot of the same sentiments as you do, but at the same time I think the popularity of the HTTP model is preventing the Internet from evolving where it needs to.

50%
50%
jheising
jheising
6/13/2014 2:00:18 PM
User Rank
Local Activator
Re: Alternative Perspective
Thanks for sharing Tim— I didn't realize that his post was so polarizing. Although I'm not sure I'd call someone posting to his/her personal Blog a "Flame War".

He responded as well: http://vasters.com/clemensv/2014/06/04/On+Discourse+And+MQTT.aspx

Anyway, I'm not trying to start a flame war myself :) I just felt that the headline was a bit sensationalistic and that there is obviously a lot of passion on both sides.

Let's just hope that out of all this comes some sort standardization, no matter what the protocol is.

50%
50%
Tim Kellog
Tim Kellog
6/13/2014 1:46:24 PM
User Rank
Blogger
50%
50%
jheising
jheising
6/13/2014 1:42:56 PM
User Rank
Local Activator
Alternative Perspective
Personally I am neither bullish or bearish on MqTT, but I think it's interesting to see a counter perspective: http://vasters.com/clemensv/2014/06/02/MQTT+An+Implementers+Perspective.aspx

I also think it's wrong to say that HTTP won't work for IoT. It's obvious it DOES work and works quite well for a lot of things. MqTT might be more future proof as networks start becoming more congested, but one has to weigh the arguments that HTTP is extremely easy to integrate with a plethora of off-the-shelf and tried-and-true products.

50%
50%
BrianHall
BrianHall
4/1/2014 3:46:33 PM
User Rank
Local Activator
Re: MQTT for beginners...
You can get access to iOS, Android and web SDKs at clearblade.com

50%
50%
kellogh
kellogh
1/19/2014 9:18:52 AM
User Rank
Local Activator
Re: MQTT for beginners...
MQTT runs on top of TCP, so it probably wouldn't work. On the other hand, MQTT-SN probably would work. it runs on datagram technologies, like UDP or SMS messages. However, I'e never heard anyone doing MQTT-SN over ZigBee and I don't know enough about it to say for sure that it would work. As for a general "getting started" guide, I wrote one here:

http://www.2lemetry.com/blog/2013/08/mqtt-in-a-nutshell

50%
50%
cpetras
cpetras
1/17/2014 6:58:12 PM
User Rank
Local Activator
MQTT for beginners...
So how do I get started with MQTT?

My application is agricultural sensors connected via zigbee/wirelessHART nodes. The nodes mesh back to a gateway/concentrator.

Would MQTT be appropriate to the node to gateway path? or is it more appropriate for the gateway to other path?

Is there any lightweigh implementations available? Or do I need to go to IBM?

 

50%
50%
kellogh
kellogh
1/17/2014 4:06:13 PM
User Rank
Local Activator
Re: Do not use HTTP
Exactly. TCP doesn't take care of reconnection on it's own, but MQTT does. On the subject of network performance, here's a paper I linked to within the text of the blog that you might find very interesting. 

http://conferences.sigcomm.org/co-next/2013/program/p303.pdf

My take on it is that TCP over cellular networks isn't a great fit. Luckily, MQTT-SN runs on datagram based protocols and is better suited for those scenarios.

50%
50%
Davidmicro
Davidmicro
1/17/2014 4:00:16 PM
User Rank
Worldwide Wizard
Re: Do not use HTTP
->Do message get delivered when the connection is unexpected broken?

I think that it may be depending on how to design the reconnection scheme. Network performance could be improved either using other method or using existing method with the updated.

50%
50%
Page 1 / 2   >   >>
More Blogs from Tim Kellogg
The Internet of Things is acquiring a language that will let people with the common touch create for the IoT.
Getting a large group of diverse IoT devices to exchange information can be a challenge -- unless they can discover one another automatically.
Twin to telemetry, telecommand involves sending data to IoT devices. The CoAP protocol is a perfect fit.
flash poll
video blogs
Inside the Internet of ...
Iain Galloway of Freescale explains what the ...

2:06

0 comments
Ingredients of the IoT: ...
In this installment of the Ingredients series, Bill ...

3:30

0 comments
Ingredients of the IoT: ...
With so many things connected to the Internet, ...

2:07

0 comments
Ingredients of the IoT: ...
Among the ingredients that make up the Internet of ...

2:22

0 comments
Ingredients of the IoT: SDN
Bill introduces the Internet of Things in this ...

2:42

0 comments
SDN's Explicit Forwarding ...
We can't prove based on current usage that SDN ...

1:52

0 comments
IoT Design & the Control Loop
You have to pay attention to the length of the ...

2:01

0 comments
Wearable Tech, IoT & ...
To make wearable tech work we need to be sure the ...

1:57

0 comments
SDN & IoT: It's Complicated
SDN and IoT have some potentially significant ...

2:00

0 comments
3 Dimensions for IoT APIs
Developers building IoT apps will need APIs that ...

2:03

0 comments
How Can We Help Stamp Out ...
What matters is not what you do in an application ...

2:10

0 comments
Another IoT View: ...
Viewing the IoT as a bunch of correlations that ...

2:16

0 comments
IoT, PnP & DIY
Smarter on-net elements for home and process ...

2:07

0 comments
Matching the Cloud & IoT
The cloud, using tools already available from ...

2:04

0 comments
IoT & the 'First Telephone ...
The solution to the IoT's bootstrap problem may be ...

2:06

0 comments
follow us on twitter
like us on facebook
IoT World
About Us     Contact Us     Help     Register     Twitter     Facebook     RSS
Video Blogs
Inside the Internet of Tomorrow Tour

10|20|14   |   2:06   |   (0) comments


Iain Galloway of Freescale explains what the IoTomorrow tour is all about.
Video Blogs
Ingredients of the IoT: Security

10|3|14   |   3:30   |   (0) comments


In this installment of the Ingredients series, Bill Kleyman talks about security.
Video Blogs
Ingredients of the IoT: Software-Defined Storage

10|3|14   |   2:07   |   (0) comments


With so many things connected to the Internet, there will be a lot of data. But where does that data live? Bill Kleyman explains.
Video Blogs
Ingredients of the IoT: Compute Layer

9|28|14   |   2:22   |   (0) comments


Among the ingredients that make up the Internet of Things, the compute layer (where the software runs) is perhaps the most important.
Video Blogs
Ingredients of the IoT: SDN

9|26|14   |   2:42   |   (0) comments


Bill introduces the Internet of Things in this first of a multi-part series.
Tom Nolle
SDN's Explicit Forwarding & IoT Security

9|12|14   |   1:52   |   (0) comments


SDN uses a central controller to decide what packets can be forwarded, so it could absolutely control access to IoT elements and provide a high level of security. The problem is scalability; we can't prove based on current usage that SDN could scale to large numbers of sensors and valid users. A second model may be needed.
Tom Nolle
IoT Design & the Control Loop

9|8|14   |   2:01   |   (0) comments


Any sensor application has a path from the sensor through a process point and back to the action point, often called a control loop. You have to pay attention to the length in milliseconds of this path or you risk getting so far behind the thing you're controlling that your application will be useless.
Tom Nolle
Wearable Tech, IoT & Authenticity

9|8|14   |   1:57   |   (0) comments


Wearable tech could be combined with IoT to create useful applications from tracking food/beverage nutrition to identifying the store with the best price by looking at various storefronts. To make it work, though, we need to be sure that what IoT provides is authentic, reliable information on which we can base decisions.
Tom Nolle
SDN & IoT: It's Complicated

9|8|14   |   2:00   |   (0) comments


SDN and IoT have some potentially significant natural symbiosis, but not necessarily in the ways everyone is talking about. Two things of special interest are the notion of a "Control and Sensor Network" built on the model of content delivery networks, and IoT security based on SDN principles, both topics to be covered in future vblogs!
Tom Nolle
3 Dimensions for IoT APIs

9|8|14   |   2:03   |   (0) comments


Developers building IoT apps will need APIs that expose IoT assets in three dimensions: geographic area, subject-tracking, and by-sensor-type. All these APIs will have to accept "credentials" so that data can be filtered by the privileges of the requestor. If all this can be done, we can count on rich IoT apps.
Tom Nolle
How Can We Help Stamp Out IoT Hype?

9|8|14   |   2:10   |   (0) comments


IoT, like most tech these days, seems to be all about hype. We can help realign things by focusing on what matters, which is not what you do in an application sense, but what technology models you expose to related IoT components and users with needs.
Tom Nolle
Another IoT View: Correlation & Context

8|27|14   |   2:16   |   (0) comments


The IoT can be viewed as a network of devices, as a web of information, or as a bunch of correlations that provide context for what we do and see. Taking this last view might help create more IoT value, guide us toward low-apple missions, and focus us on the real security concerns.
Tom Nolle
IoT, PnP & DIY

8|20|14   |   2:07   |   (0) comments


Traditional sensor networks don't put low-level elements on the Internet for cost reasons, but smarter on-net elements for home and process control could make the IoT a lot more accessible. In the end, it may be worth the extra cost to get extra participation.
Tom Nolle
Matching the Cloud & IoT

8|15|14   |   2:04   |   (0) comments


The cloud, using tools already available from providers, could offer significant value-add in IoT applications, particularly in making data available from a large, diverse set of sensors and turning that data into big-data/Hadoop repositories or workflows.
Tom Nolle
IoT & the 'First Telephone Problem'

8|15|14   |   2:06   |   (0) comments


You can't sell the first phone, the saying goes, because there's no one to call. The IoT may face a similar bootstrap problem. The solution may be to focus on placing sensors on the objects that interest us, adding environmental sensors as we build the business case.