Latency and reliability

[timb]

I have been working with z-wave.me now for just over a year. I love the concept, really want it to work, but am frustrated by the overall reliability and performance of the system. I am not pointing fingers at the software but the whole z-wave system. I would like to share my experience and see if anyone can give me pointers to make improvements. The network I have is as follows:
18 devices in total:
• Fibaro Dual Relays x 3
• Qubino Dual Flush relays x 2
• Fibaro Multi-Sensor x 2
• Philio Dual relay x 1
• Fibaro Universal Binary Sensor x 3
• Everspring SP103 PIR
• Plug Switches TZ68 x 3
• Z-Wave.me Remote Control
• Aoen Labs Multisensor 6
• Pi 2 Controller with Z-Wave chip: ZW0500
Software:
• RaZberry by Z-Wave.Me
• V2.2.0
• Serial API 05.04

HARDWARE RELIABILITY
In the course of 12 months I have had 2 Fibaro Dual Flush relays fail. I replaced them with Qubino Dual Flush relays. Within 3 months, one of those devices seems to be on the edge of failing. It appeared dead, I tried to exclude then re-include, but only 30% of the points of interview have worked the rest remain un-interviewed after countless attempts. I fear another replacement is due.

A failure rate of 3 out of 20 devices is 15% within 12 months. This is a crazy figure. In a hobbyist environment like mine you might live with it (although you should not!), but this kind of reliability would never make for commercial deployment.

LOSS OF CONNECTION
On a monthly basis one or two devices seem to lose their interview connection and I am forced to re-interview. Is this a hardware issue of a software issue? I don’t know. It is frustrating and time consuming for me to have to repair the network with such frequency.

INTERVIEWS
A number of my devices have always failed to interview completely:
• Fibaro Universal Binary Sensor – fails with Alarm Sensor
• Qubino Dual Flush relays – multiple items failed: AssociationGroupInformation, MultiChannelAssociateion, Version

I don’t know what kind of impact this has, but I can’t fix it.

SOFTWARE
I have lived through installing the Z-Wave.me software over numerous iterations from v1 to v2.2.0 I learned the hard way not to install “RC” versions, which on several occasions caused me to lose the whole set up and I had to start again. Right now I find v2.2 quite stable and mostly bug free. It has improved greatly from the early days, well done to the team for all your efforts – please keep going!

However I do get frustrated that the Expert UI seems to regularly forget the selected Device Description for many of my devices.
Also I really wanted to do Sensor Value Logging, but in the end turned this App off as it seemed to make my whole system perform very badly.

LATENCY
Much of what I want to do is quite simple like to turn on lights when I enter a room. I use existing burglar alarm PIR sensors wired to the Fibaro Universal binary sensor as inputs and then the relays to turn on the lights. I do all the logic in the Pi Controller using Rules and Scenes. I know I could associate devices directly but then I don’t get the added control I can have in my logic, for example only when I want to go into Auto Mode, only when it is dark….
The problem is that the delay between triggering the PIR and turning the lights on is long and variable. The best is about 1 second, normally around 2-3 seconds, but sometimes a whole minute can go by. It means I walk into a room, in the dark and I have walked out the other side, before the lights come on behind me. Again not practical for commercial use. I have two questions:
• What is the cause of the latency (I monitor my Pi CPU and it is less than 4%)?
• Why does the latency sometimes get very extended?

SUMMARY
Overall the picture builds to a system that is unpredictable, unreliable and requires regular maintenance. Don’t get me wrong, I LOVE the idea and I REALLY want it to work. I have invested a good deal of time and spent a good sum of money. It just doesn’t work well enough.
Am I alone? Do others have the same problems? Is this the best I can expect? I would really like to hear the Developer’s opinions. Any pointers would be gratefully received.

[jet11x]

Very well put.

I have a smaller setup than you and have had a similar experience except:
1. I've never lost an interview
2. The Expert UI has never forgotten a device description
3. I've never noticed latency above 1-2 seconds
4. I had to give up for now on v2.2 as I lost cron/scheduling
5. None of my devices have failed but I've only had 4 for over a year

I'm going to stick with Razberry and am still adding sensors - I hope the software quality will continue to improve.

[timb]

Does anyone have any idea what could be the cause of the latency and is there anything I can do about it?

[pz1]

I can't help you with the latency since I do not have motion detectors. But for the most part your analysis, given in the first post, is spot on!
I have expressed similar concerns to the Z-Wave alliance, but they prefer to remain silent That is a pity, because overall I still do like the technology.

[timb]

Thank you for replying, I appreciate it. Otherwise I feel is it only me!
Some more info....

I have the zwave.me remote control - when I press a button to activate a switch there is minimal latency. But that device has a direct association with the switch, so I assume it does not go through the Pi/software. That I think eliminates the z-wave rf transmission from the equation.

As for the PIRs, when they sense motion they illuminate a LED on the PIR so I can see when they detect. I am measuring the latency from the point in time the LED activates to the time the lights switch on. The PIRs are hard-wired into the Fibaro binary interface. So now when the PIR activates the events should be:
Binary Interface detects motion - setting the sensor value ON
There is a Rule app that says "if motion ON then activate scene turn lights on"
Its simple.
Like I say sometimes it works within a second or two, but other times it can take much longer. It's like the Pi is busy/overloaded. I do have something like 20 scenes and 15 rules - is that big by anyone's standards?
The thing is, when I monitor the CPU it is pretty much always close to zero, the very occasional peak at 24%.

Along with this, sometimes (1 time in ~20) the UI interface seems to be slow to respond to a click. I can understand that if I ask it to load the Event Log, but examples would be to navigate to Active Apps. Usually fairly instantaneous, other times I wait longer. It feels symptomatic of the issue above. Its like the Pi is busy and locked up for a bit. But again - no evidence on the CPU usage.

Regarding the CPU usage, the Pi is a Quad core, so I assume 25% is one core maxed out. Is the z-wave software multi-threaded? Could it be it is single threaded and one CPU is performing some background task, maxing it out and I just have to wait?

Perhaps you can point me in the direction of how I might be able to investigate further. I can see the log file in /var/log/z-way-server.log But I have to say its not easy to figure out what is going on. Any pointers?

I really like this stuff and really want it to work. Just frustrated that I can't get to the bottom of the issue and its holding me back from investing more.

Thanks.

[pz1]

The only thing I can suggest as a wild guess, is to look at the Z-Wave queue. (You can select that via ExpertUI, Controller Info. Push the topleft button on that screen.) See if there is a very long queue, or if the queue is stuck (jobs don't disappear)

[marco]

Hi timb,
for now, I'm just writing to tell you that you are not the only one with this kind of problems.
I hope to come back quickly with my experience.
Bye
Marco

[pierre2302]

is that in the "timing info" of the expert interface there is not a module that includes lots of red number ?

[timb]

HI all,
Thank you for your posts, sorry it has taken me a while to reply.
A couple of points to update you...
1. I moved the location of my Pi - hoping to get better transmission - there was some improvement.
2. The timing info shows mostly green (direct connection), some black connection and the occasional red - on the furthest devices. I will continue my investigations at the moment the Jury is still out.
3. I had another failure of a switch - this time a Qubino dual relay. This device is switching 6 LED lamps - the 5W high luminosity variety. I had read the contacts can arch when disconnecting the power causing the contacts to weld together. That seems how it failed - the switch was stuck ON. I had even wired in two anti-arch suppressors. Does anyone else have issues like this? It seems these 6 LEDs have taken out my Qubine dual relay and the FIbaro before that.
Thanks
Tim

[lvmeijer]

I'm having a Zipato Zipabox controller and have 60+ Z-Wave devices in my network. I'm having the same trustrations as you do. Especially my Qubino devices have issues. Sometimes they disappear in the network and become unusable. Somes I I swap the fuses off/on in the house, they appear in the network again. And I'm also seeing latency issues... I'm not sure if it's the controller or the network.