How to choose a server and location

The server is the foundation, and it's where people most often get it wrong before the first command. Let's work through the selection logic: location, host, IP whiteness, spreading across networks. Without it, any perfect config sits on a rotten foundation.

The main rule of 2026: exit only abroad

Let me start with what defines everything else. Exit nodes — the ones through which client traffic actually flows to the internet — are kept only abroad. Not because it's "freer" there, but because the regulator sends Russian hosts lists of VPN-server IPs and demands blocks. There are already precedents: providers get a subnet and give you a day to remove the service.

Hence the split of roles by geography:

• Exit nodes — abroad. Low-ping-to-Russia locations: Germany, Finland, the Netherlands. Poland also gives good ping and a white exit.
• Panel, bot, relay / cascade entry point — can be in Russia. No client exit traffic flows through them, so there's no abuse risk.

If you put an exit node on a Russian host, the question isn't "will it get blocked" but "when." And on top of that, someone else's traffic will flow through your exit — and the server owner will answer for it.

Different providers and different ASNs

The second rule beginners ignore and then lose the whole fleet overnight: nodes must live in different ASNs (autonomous systems — roughly, different providers and different subnets).

The reason is that blocking is often done not by a single IP but by a whole subnet. If all your nodes are with one host at adjacent addresses, they get wiped out in one move. If they're scattered across different networks, they fall one at a time rather than all at once, and you have time to migrate. It's cheap insurance that will one day save the whole service.

IP whiteness: the silent killer

Here's where half of beginners burn out. They grab a "cheap" IP, layer a perfect Reality on it — and it works every other time. The cause is a dirty subnet: the address has been on blacklists for six months because someone made noise on it before you, or it's a known proxy subnet that filters recognize on sight.

The key thing to understand: IP whiteness is decided before production, not after. You need to check the address before putting it into service:

• not on blacklists, not in a flagged proxy subnet;
• reachable from Russian mobile networks, not just from home internet — mobile operators have stricter inspection, and an IP that pings from home may not work on mobile under an "allowlist."

A true white IP is exactly the one that gets through even mobile "jammers" while also not sitting on blacklists. That's an expensive and valuable resource; you gather them separately and guard them. Sourcing white addresses for the right subnets is a big topic of its own (for this we have a tool called MultiRoller), but remember the principle right away: a dirty IP kills any config.

Resources: how much hardware you actually need

This is where beginners overpay. The requirements are modest:

• Entry/exit node — 1 vCPU and 1–2 GB RAM handle dozens to hundreds of clients. Xray is undemanding.
• Panel — a separate VPS, 1–2 vCPU and 2 GB RAM. Put the panel on 2 GB with headroom: on 1 GB it starts with swap and risks running out of memory.
• Iron rule: panel and nodes go on different machines and different IPs. Never keep the brain and the traffic on one address.

A single node on 1–2 GB serves many clients, so as you grow, the per-client cost falls — you scale not by amount of hardware but by smart distribution.

Hosts: what to look at

A "best top forever" is impossible — the market shifts, and what's alive today is under a block tomorrow. So first, the criteria to choose by yourself:

• For exit — a foreign host whose subnets show up in block lists less often. Niche European providers are often better than mass-market ones precisely because their networks aren't in the spotlight.
• For Russian infrastructure (panel, relay, bot) — a provider with ruble payment and hourly billing, so you can shut down excess without losses.
• Payment and anonymity — wherever possible, crypto or SBP without extra ties to your identity.
• Never your only node with one host. Even a good provider has sudden blocks and outages. Always keep a reserve in a different AS.

And so you don't start from scratch — the ones I've run myself or keep on my radar. This isn't an ad or a forever-top, just a starting point: check they're still relevant and read fresh reviews for your case.

• For exit (abroad): niche Europeans are usually whiter than mass-market ones. I look at Netcup and Contabo (Germany), PQ.hosting (multi-location, takes crypto), Zomro (Netherlands), 4VPS. Big Hetzner has subnets long known to the filters, so I take it for the panel and tests rather than a white exit.
• For the Russian part (panel, relay, bot): Selectel, Timeweb, RuVDS, FirstVDS — rubles, hourly billing, quick start. Beget works fine for light Russian entries.
• Anonymous payment: where possible, crypto; of the above, PQ.hosting and some of the Europeans accept it.

Mixed reviews for a host are normal in this niche. The question isn't "is it perfect" but "do I have a backup if it lets me down."

Hygiene that saves money

A couple of things that aren't about tech but about your wallet:

• Delete unnecessary and stopped VPSes. An idle server keeps ticking on the bill — it's the main silent killer of margin.
• Monitor your hosting balances. Hit zero and the server gets shut down, clients are left without a channel, and you find out from the stream of negativity.
• Secrets — only in a secure place, never in plaintext in a repository.

What's next

Server picked, roles separated, IPs white — you have a foundation. Next in this section is the start checklist: concrete verification steps before you stand up your first server — this time with commands. And after it — installing the panel, which we'll hook this server up to. Remember the main thing: a perfect config on a rotten foundation still falls, so choosing the server isn't a formality but the first real decision.