I use Welogic 12c and Coherence? coherehce cache broadcast to every coherence node on the network? how to limit broadcasting to certain addresses?
you should use Well-Known Addresses (WKA) feature instead of multicast option.
See answer to the similar question Here
12c documentation Here
You can use WKA for discovery to eliminate multicast from discovery.
You can also disable multicast altogether for communication. Coherence does not require multicast.
Related
I am quite new to DDS. I am using ROS2 which is built on DDS. I want to run a ROS2 based publisher and subscribe to a RTI Connext based Simulink model both located on different PC's.
I want communication between 2 different PC's.
Questions :
Should both PC's be located in the same network.
Should multicasting be enabled ? Multicasting is blocked in my company's network. Is there an alternative to this ?
Yes, both PC must be under the same network.
Multicasting is used for the discovery of the nodes in ROS2.
If your company does not provide the necessary networking support, you can always try to build your own. For example you can use a switch and connect both PC to it.
Or you can use your phone. I created a hotspot with my phone with 4 devices connected to it and communicating using ROS2.
However, if you need to have access to the company network, ask your IT department if they can set you up with a switch connected to the network that has multicast :)
Edit: As specified in another answer, it is possible to do a static discovery. However, this would imply modifying the ROS2 configuration of the RTI Connext RMW> https://github.com/ros2/rmw_connext
If the computers are on the same network, and multicast is enabled, and the multicast TTL is high enough, and the computers are publishing/subscribing on the same topic on the same domain, and their Types and QoS are compatible, then they should automatically discover each other.
If multicast is disabled, you can tell the DDS participant to use a specific address to peer to (see the docs for "initial peers"). This can be done in different ways. Here's a good reference: Initial Peers
I would like to do it with the Reliability of a TCP Unicast.
I have used UPnP based Multicast (M-Search) but many devices filter out Multicast messages so I end up losing them.
Also, how does Bonjour compare to UPnP?
For the second question
: In aspect of the functionality and reliability, uPnP and Bonjour are very similiar. The main difference of them is how prescriptive is. uPnP is more prescriptive.
It's my understanding that IP multicasting works by receivers joining IP multicast groups with IGMP messages, senders sending to those addresses, and routers addressed in that way duplicating the message and sending to all receivers in the group.
I've never elected to join a multicast group. What software that I might use would likely join a group on my behalf? Do commonly-used programs often send to multicast addresses on my behalf? What tools, especially command-line tools, expose this information? Something similar to what ipcs provides for shared memory, what ps provides for processes, etc. I'm especially interested in linux or OSX environments.
Use
netstat -g
NB there is no duplication. If a router has members inside, it propagates the datagram, once, on the inside LAN.
I am learning about MPLS VPN networks. From my understanding an IGP runs on all core routers (P and PE), while BGP runs on all PE routers. Once the IGP has conveyed reachability information to all routers, and all routers have converged, the exact labels to be used to transfer packets are assigned using LDP.
My doubt is, how are BGP control packets transmitted between PEs.
There are two options.
1. To use the Label switched paths between PEs.
2. To use normal IP forwarding through the P routers.
Which of these two methods is actually used?
If both can be used how does the PE router make a decision on which one to use?
Do we have to manually configure it in the router?
Do these answers vary for different routers like Juniper, Cisco etc. ?
My doubt is, how are BGP control packets transmitted between PEs.
There are two options:
To use the Label switched paths between PEs.
To use normal IP forwarding through the P routers.
If both can be used how does the PE router make a decision on which one to use?
LSPs are preferred over per-hop IP forwarding, if an LSP is available.
Do we have to manually configure it in the router?
By 'it', do you mean configure use of the LSP for BGP control-plane information? It happens automatically on a Cisco IOS box
Do these answers vary for different routers like Juniper, Cisco etc. ?
Cisco will send BGP information through an LDP LSP, as long as the BGP endpoint prefix has an label binding.
I can't remember Juniper's behavior off-hand, they offer somewhat more granular control over LSP behavior.
BGP uses TCP te setup its connection and to send their packets to his neighbors.
This means that your neighbors need to see each other on layer 3 (ip) level.
I hope this is the info that you needed.
see: http://en.wikipedia.org/wiki/Border_Gateway_Protocol
section "operation" for more details on this matter.
I am designing a system which has multiple small embedded systems hosts communicating in a LAN. UDP multicast fits nicely for my purpose.
But I am worried about multicast support in commonplace routers. I need to deploy the system to normal households equipped with a Wifi router, so I could encounter any kind of routers. I will use UDP broadcast if multicast will be more trouble than benefit.
To decide, I am thankful for any data or experience on multicast support in today's commonplace routers:
Do all consumer routers sold today support multicast reasonably? Limitation to LAN is not an issue for me, I do not need multicast across the Internet.
How about older routers?
Are there any big issues in commonplace multicast implementations I need to be aware of (e.g. packet drops, configuration issues, etc.)?
Are you talking switches or routers? In a consumer setting I suspect switches. My experience is that they all support multicast, though not at wire speed. Also the cheap ones tend to broadcast any multicast traffic to all ports (no IGMP snooping). Packet loss is definitely something you need to deal with, it can and will happen even on 'professional' networking gear.
Edit: as long as you are in switched network, you typically don't need to configure anything.
In many scenarios the equipment either does not support IGMP spoofing, or it is off by default. There are two problems:
Any wireless interface can be saturated by the traffic.
Poorly configured units may inadvertently route traffic out the default gateway stalling legitimate traffic.
In either case, you will have your equipment discarded as the cost of investigation will almost certainly outweigh the benefit received.
If your traffic has a limited rate and you are not concerned about the WiFi impact, you could use the local broadcast address to ensure delivery to recipients, without impacting on the routed communications.
You could install a discovery mechanism into your nodes it may be worthwhile to implement a unicast overlay to ensure that traffic does not have inadvertent impact.
A single larger group of customers with some non-compliant devices requesting support will swamp any costs associated with development or additional traffic caused by not implementing true multicasting.
In worst case scenarios, when the routers don't allow multicast traffic, I would encapsulate the multicast packets into a unicast IP address. This way the routers would handle them as normal unicast data. You might want to check mrouted .
Good luck