One of my readers recently pointed me to a blog post written by Andrew Lerner from Gartner describing the drawbacks of stretched VLANs.
Andrew starts by listing two misconceptions…
- That we need stretched VLANs for VM migration (check);
- That we need stretched VLANs for active-active data centers (check).
… and lists four things to consider:
- WAN latency in LAN environment (check);
- L4-7 service architecture and resulting traffic trombones (check);
- Resiliency and split-brain (check);
- Traffic symmetry (check);
Finally, he concludes that “While these approaches may appear as a simple fix, they often increase complexity, reduce performance and availability, and increase the overall cost of your network.” Yep.
Why am I writing about this?
While it’s nice to see someone else coming to the same conclusions (particularly against strong headwind of vendor marketing), what matters most is that “Gartner Said So”.
I’ve heard sad stories from dozens of engineers telling me how they tried (and failed) to use my arguments to persuade their bosses that we shouldn’t overcomplicate the network to solve other people’s problems – now they can point to a Gartner report saying exactly the same thing.
Unfortunately, for some of those engineers the report came years too late – at least a few (that I know of) already got badly burned by a meltdown spreading across multiple data centers.