You are here

DHCP "HOST" Settings 1H vs 5H vs 13H

10 posts / 0 new
Last post
k1ky
k1ky's picture
DHCP "HOST" Settings 1H vs 5H vs 13H
I am looking at streamlining our node operations to cut down on overhead since our system has and is growing.  In the past, I have almost always set new nodes up as 13 Host - just to stake my territory..... Now that I'm a little bit wiser, I'm re-thinking that strategy.

For nodes that don't and will not have any external devices like phones, cameras, computers and the like I don't see the need for anything more than 1 Host.
Nodes that may have a phone, camera and a computer will now be set to 5 host.  At some of my sites I may have 6 nodes or so operational - is there even a need to have DHCP running on some of the "Relay" stations?

13 Host would be used rarely if ever.

Is there any data on how these setting affect memory usage and OLSR processes?

Is this a "sound" strategy for node deployment settings??


 
AE6XE
AE6XE's picture
AREDN address space and chance of an IP conflict

Conserving resources is always a wise strategy.   The node storing in memory the definition of a subnet, regardless if 1, 5, or 13 devices, is still the same memory footprint.  For example,  this would look like 10.11.12.1/30, 10.11.12.1/29, or 10.11.12.1/28.   There's not an option in the UI to turn off the LAN subnet and free up the memory.  We'd need to add an option to enable this in a future firmware version.    Turning off DCHP would save a few memory bytes on the one node.  

The issue with th 1, 5, and 13 addresses and limiting the count is to avoid an IP address conflict.   However, if one were to ever occur, a manual correction readily resolves.   To put into perspective, we have 2^24 bits of address space in 10.x.x.x.    This is 16,777,216 possible addresses.   Each node would consume:

WAN - 1 Address
DtDLink - 1 Address
LAN -  4, 8, or 16 Addresses (for 1, 5, and 13 devices)

Thus we can have this many mesh nodes to fully consume the address space: 

if everyone had 13 LAN devices - 932,067 max mesh nodes
if everyone had 5 LAN devices   -  1,677,721 max mesh nodes
if everyone had 1 LAN device     -  2,796,202 max mesh nodes

We'd have to known how Ubiquiti assigns MAC addresses with the 4 or 5 manufacture #s assigned to them to determine the probability of an IP conflict.  But we'd likely need to see a contiguous mesh network reach 100,000s of mesh nodes before there's any meaningful risk of an IP address conflict.

Joe AE6XE

k1ky
k1ky's picture
Nice description Joe
Joe, Thank you for the well crafted response to my query.  Probably worthy of including in the node HELP file, which I have read on many occasions.  One thing you didn't touch on was what effect the larger 5H and 13H settings have on the OLSR traffic and actual data usage per node. Any insight on that? Significant or no?  We have some tunnel users who are using limited data plans and I sometimes may use a mobile hotspot to tunnel into our system from remote sites and I'm looking to reduce unnecessary tunnel traffic as much as possible.  It's looking like about 8Mb per hour usage idling on our system.


 
AE6XE
AE6XE's picture
I'm not seeing any effect on
I'm not seeing any effect on OLSR traffic.   The subnet is advertised and propagated by OLSR and would look like one of the follow:  10.11.12.1/30, 10.11.12.1/29, or 10.11.12.1/28.  (The '30' is 1 LAN device, the "29" is 5 lan devices, ... )

Joe AE6XE
N2MH
N2MH's picture
Actual Advertised Services

Joe,

Correct me if I'm wrong, but the OSLR traffic is mostly affected by the number of advertised services and the information along with those services (ip address, service name, url if applicable, etc.). Thus, in the context of saving OLSR traffic, any unused advertised services should be deleted and no longer advertised. I've seen in a typical nodes list that many advertised services are unreachable, probably because the machine behind that is off line, sometimes for months on end.

To Tom's point, something to keep in mind: If you do downsize (or change size either way), any services or hosts on that node will get a new ip address due to the algorithm that generates the DHCP address range. Thus, if you advertise phones with an ip address so that people can reach them by direct ip dialing, those advertisements will need to change. If the phone only has to register with a pbx, no change has to be made. In fact, the phone does not even have to advertise itself if no direct ip dialing is involved. Only the pbx has to know the ip address of the phone and that is done automatically when the phone registers itself.

Mark
 

AE6XE
AE6XE's picture
Yes, each advertised service
Yes, each advertised service sends bytes around the mesh every few seconds.   But let's put this in perspective, we're talking  less than 100 bytes for the typical service advertisement.   While every bit counts, this isn't the first place I'd be looking to cut down on traffic.   For example, you might check everyone's laptops on the network and see how many of them are constantly trying to reach out to do windows updates, dropbox sync, one drive sync, and the list goes on.  Many networks have internet gateways, so unbeknown to the owner, it's working to download that latest Gigabyte update.       Most every ipCam is continually trying to reach out to the internet for the iPhone/Android apps to connect back to.   There's bigger fish to fry... 

Joe AE6XE
gm7gde
DHCP pool has changed subnets
I have a Uqbuiti Air node that has been running for months now, but something happend last night and as a result I have a new IP pool, which is a big issue now as I have remote Cisco phones hardwired with the IP of the tftp server on my node which no longer has that old IP address.

I haven't seen this before, only when you move from 1 to 5 or 13 Host but I am still running the option of 13 Host, question is what perameters generate IP subnet for LAN and what could I have changed that generated a new subnet ? Can a manual configure it with the old subnet ?

Regards

Andy
K5DLQ
K5DLQ's picture
did the addresses change on
did the addresses change on the same unit, or, did you replace it?
AE6XE
AE6XE's picture
How IP addresses are determined
Andy,  

Check out this document:  http://bloodhound.aredn.org/products/AREDN/wiki/TechRef/GUI/admin/PerlUI...

Best practices I've used in the past --  configure applications to reference hostnames and not IP addresses.  This way, if you have to replace a mesh node due to hardware failure or otherwise, the IP address can/will change.  The same hostname can be assigned = everything still works.

Since the IP addresses trigger off of the MAC addresses, something must have changed in the config files to reset to 5 addresses and back?   Holding reset button can do this?  restore defaults.   We've even seen water damage shorting the cat5 wires which ubiquiti uses to do the remote reset button.  

Joe AE6XE
gm7gde
didn't change anything
The config is still there as its also running as a Tunnel Server and all the tunnel clients are still configured along with the reserved attached devices using their MAC but now they are getting new IPs.

I am baffled by this before it was 10.4.11.18/28 now its 10.37.16.161 / 28

same hardware all that happend my clinet PC no longer was able to bring up the page so I restarted the PC and it got a new IP and I was able to connect to find the IP had changed.

Not all device can accept host name as the cisco phones have an IP set for tftp address and I will now need to remotly connect to all the phone and update their configs

still would like to figure out what happend, I don't have a note of the MAC before incase some way this changed

Andy

Theme by Danetsoft and Danang Probo Sayekti inspired by Maksimer