Needed Features for Network Topology Mapping
We need the ability to create topology map of our switches and routers showing the actual links between interfaces on the devices and showing the latest InMbps / OutMbps datapoints for each interface...
Forum Discussion
Stuart_Weenig
Legend
LegendI have not gotten LM topology to approach the utility of other products' topology capabilities (like CA Spectrum).
Mike_Rodrigues
Product Manager
Product ManagerOther than what you've already laid out, are there any other capabilities you'd want to see in LM Topology to make it more comparable to something like CA Spectrum?
- Stuart_Weenig
The main reason we don't rely too much on mapping is because it's hit or miss. It may have some of the relationships, but i couldn't say with more than 60% confidence that the map is complete, even from a L2/L3 perspective, even if I knew all the devices were in LM.
We need OOTB TSs that link all instances to their devices. Actually, it would be best if instances were related to their instance group, which is related to the datasource, which is related to the device, which is related to all the groups they're in. Basically the resource tree in topology form. LM has all that data already, just need to allow instance groups, datasources, and resoruce groups as nodes.
That would go a long way since it would meet the use case of showing the device connected to the interface connected to another interface connected to another device.
Like give me 4-10x the number of topology sources than you have right now. You'd get close to the 30+ years Spectrum has building their topology sources.
- an OOTB TS that connects devices to their assigned collectors for better RCA (I forget the new name).
- Device -> IPSEC tunnel -> Interface
Pie in the sky idea (if LM really wants to be powered by AI):
Let me draw a connection between two objects and have AI suggest a topology source that would automatically map other objects that are similar. The AI would at first suggest 10 more pairs of objects it thinks should be mapped together. I'd tell it which pairs it got right and which it got wrong (and which i'm not sure of). Repeat enough times to get enough "yes this pair of objects belong together" responses to train the AI. Then pop out the TS.
Prompt me every once in a while to continue training the AI the same way Google Photos prompts me "Are these the same person?". Just pop up a notification that asks to help the AI get better. When I click the notification it presents me a series of pairs of objects that it thinks are the same. I click yes/no/idk on each one, which trains the AI, which then tunes the TopoSource.
neilticktinNeophyte
Mike_RodriguesFor us, like Stuart, (but not anywhere near as sophisticated) the biggest problem is just knowing that we have everything mapped, or called out that it's not mapped ... and for the MAC addresses to just match up. For us, the MAC's could come from cdp, lldp, or EVEN a csv that we can just set a property for. It's rather a headache to not have the simplest things to depend on or easily upload.
