Vancouver here. I use Vultr's Silicon Valley location for work stuff because clients expect west coast speed. But I noticed they opened Seattle - has anyone tested latency from Vancouver to Seattle vs SV?
My informal ping says 18ms to Seattle, 35ms to SV. Seems obvious but wondering if routing is clean.
From Vancouver that's expected. CIRA's looking glass shows clean paths to Seattle via Seattle Internet Exchange. Vultr's Seattle is in their Westin building presence, same as many others — https://www.peeringdb.com
For your use case Seattle saves you 17ms but Silicon Valley has more peerering diversity. If it's just client-facing static content, Seattle wins.
Member
Melbourne to Vultr's Sydney: 22ms stable. To their Singapore: 85ms. To Amsterdam: 320ms.
The Sydney location is fine but expensive compared to their US tiers. I run personal stuff on US locations and accept the latency. $5 vs $10 for same spec.
OVHcloud's Sydney is cheaper but the routing to Melbourne goes via Perth sometimes. 45ms instead of 22. No thanks.
Member
OVHcloud's Asia routing is weird in general. From Taipei their Singapore location routes through Hong Kong first. Extra 15ms for no reason. Their network map says direct but reality disagrees.
I use Vultr Taipei for low latency local stuff, HostHatch Singapore for storage. WireGuard between them, best of both. I used https://www.wireguard.com for the tunnel.
WireGuard is fine but have you considered the implications for your threat model? Not the encryption, the endpoint discovery. Both providers now have your traffic pattern.
I run my own WireGuard endpoint on a Hetzner VM in Falkenstein. Then mesh to wherever. One point of trust, not two.
This is the container skeptic speaking: you're just moving the trust point. Hetzner sees your traffic instead of two others. And now you're maintaining a VM for it.
Before Kubernetes we called this "a server." It was simpler and we were happier.
Correct. But I choose to trust one German privacy law instead of two US cloud terms of service. This is a legal threat model, not a technical one.
Also I run ZFS on that VM. Send me your thoughts. I have time.
A VM. With ZFS. The virtual disk is probably already on some SAN with its own redundancy. You're adding layers for no reason. Just use ext4 and snapshot the hypervisor.
Unless you need compression. Then I forgive you. Barely.
Compression and send/recv to my home NAS. The VM is just a relay point. ZFS lets me snapshot, compress, and incrementally send. Try doing that cleanly with ext4 and LVM.
This thread was about negotiating with providers. How did we get here.
Every thread about VPS eventually becomes about filesystems or WireGuard. It's a law of nature.
Back on topic: I tested the "competitor price" technique with KnownHost. Mentioned RackNerd's pricing for similar specs. They came back with "we're managed, they're not, but here's 15% off your first quarter."
So the technique works even on managed providers, but the discount is smaller and time-limited.
