i think the reason i chose to have a separate core is to have dedicated redundant hardware as L3 core to serve access swithces as well as have link to Internet firewall and another link to SD-WAN Vello connecting the site to other sites.
If your thinking you're best served by the logical and physical topologies adopted, that's fine! Whether that's actually true, I cannot judge; I have insufficient information. However, again, good designs probably deserve more consideration beyond greatgrandpop, grandpop, and pop all did it a particular way, automatically make it the best way now, doesn't make it true. The converse is true too, i.e. a new approach isn't necessarily better either. Just consider what's actually best for you.
Whats the deal with not spanning same vlan across L2 HW?!
It continues to create potential L2 issues that are avoided by using L3.
Consider L2 hubs vs. L2 switches. Both provide L2 domains. Both move frames between hosts. So, why did practically everyone migrate from hubs to switches?
Just as switches are superior to hubs, L3 switches are superior to L2 switches. I believe some reasons why L3 switches haven't totally replaced L2 switches include reasons like: cost; the advantages don't seem as impressive; it seems to take more work to set them up, especially compared to the way we're accustomed to using L2 switches.
management vlan and voice vlan both of them needed everywhere ......
But do they all need to be in the same physical or logical subnet/VLAN?
In my last job, I had a couple of L3 6513s using 96 FE PoE port line cards, each port supporting a voice and data VLAN. Across a single chassis, 11 time 96 equals 1,056 edge ports. Do you think I had just one voice and data VLAN on that chassis, and spanned those two VLANs to another 6513 equipped the same way?
I vaguely recall, those two 6513s connected to a 6504, which hosted two off site links. Why use a 6504, couldn't the two 6513s connect to each other and each have one of the off-site links?
Yes, one aspect was the (questionable) hope if one edge 6513 went nuts, the 6504 protected the off-site ring from injected instability.
However, as OSPF was the routing protocol being used, to ECMP the off-site links, they had to be on just one device.
For either both reasons, a 6504 with dual sups is an expensive solution, but the 2,100 edge ports was supporting a call center, so avoiding a major outage was high priority. So, a good design takes into account many factors.
At the time, the 6513s, would have been replaced with newer technology, perhaps using the (then) new VSS technology (which we were starting to use). (Also, at that time, we also had one facility migrate to using new Catalyst 6xxx IA approach, which eventually allowed about 4,000 edge ports per one L3 switch chassis.)
If done today, possibly a spine/leaf might be considered.
