TierHive is a no-SLA, budget service, probably the cheapest generally available KVM VPS that exists online right now, run by experienced, competent people. But it is something new, it's a new way of doing things. It is trying to fill the gap between a homelab, a local VPS running on your own desktop, and enterprise or aspiring-enterprise hosting. It is not trying to be enterprise, it never will be. It's meant to be an ultra-cheap budget service, running on older hardware to achieve that price point. It is not fast, and it is not the place to host things you demand uptime for. If things go wrong, and they will from time to time, we are going to do our best to bring everything back online, but ultimately it may not be quick to recover if we lose a node or a location.
We provide HAProxy and failover, and every tool you could need to make your stack redundant on TierHive. If you choose not to, to save a few pennies (literally), that is a risk you take. If you put data on TierHive, we do everything possible to avoid any data loss, but in the event it happens, we will not recover it for you, you get a fresh, blank VPS. Your data is your data, because it is yours. If it is important, have a backup, and a backup of that backup.
TierHive works because we are using decades of experience and tuning to get maximum density out of our hypervisors. Everything in TierHive is custom, we designed and wrote it on top of an absolute minimal, cut-to-the-bone QEMU/KVM/libvirt stack. We constantly change things to tweak and tune them for maximum density and performance.
Our goal is usable for everyone, not fast for one person. If you just see the price tag and want to load 100 x 4GB VPS instances running Minecraft bots or whatever because it's cheaper than DigitalOcean, you are probably going to have a bad time.
We limit disk IOPS, CPU, and network.
As each location has slightly different hardware and limits of its own, it's hard to give uniform figures. Forcing uniform hardware would give us unfavourable pricing, and we would have to charge more. A server in Singapore does not come with the same network resources or price as the same server in France, that's just how the world works. We have done our best to balance this out, but it does result in some inconsistency between locations.
We have 3 tiers: low, medium, and high. In most cases, as standard, low = 100, medium = 250, and high = 500. We also configure this with burst buckets, so you get 5 seconds of meaningful burst at 2x or 4x the limit on demand, meaning real-world use will be higher than these limits on demand/burst, so database writes can happen fast, and other similar short-lived I/O can happen faster on demand. Obviously, synthetic benchmarks like fio/dd tests are going to get hard-limited by this, by design. If you need constant, reliable, high IOPS, TierHive is the wrong service for you, we can't provide that at the prices we charge.
If you need dedicated cores, please do not use TierHive. We allow access to 1 vCPU for 2GB and under, anything over 2GB gets access to 2 vCPU. We do not guarantee any specific CPU model, but we do have passthrough and nesting enabled for you, in case you want to run anything that requires it. CPU resources are by far the most contended, you should expect to see higher steal levels than you would see with any enterprise-grade host, we do what we can to reduce the real-world impact.
Steal is often misquoted as meaning if you see 20% steal, then you're being slowed down 20%. That is not what it means: steal is the percentage of time your virtual CPU was ready to run but had to wait for the physical CPU, because another VPS was using it at that moment. It doesn't translate directly into your workload running 20% slower, short bursts of steal are usually not noticeable in normal use.
We have 3 tiers for CPU: low, medium, and high. These are simply different CPU priority tiers through weighting on the backend, meaning if you are on a low tier and someone else is on a high tier, and there is contention, they will get cycles faster than you.
We have a CPU watchdog: if you average 100% of a CPU core for more than 2 hours solid, you will be limited to 50% of a core on the backend. You can self-resolve this with a reboot if you are limited.
We do not give any specific arbitrary GB or TB number, it is simply not practical with the TierHive model. Instead, we do performance tiers mixed with a FUP/QoS system.
TierHive has 3 network tiers: low, medium, and high. The maximum available speeds per location are on the TierHive homepage, and also available when creating a VPS. We also allow burst, with 2x the advertised speed for 10 seconds, so if you see 100 Mbit/s, you will be able to get 200 Mbit/s for a short period of time, which helps with real-world activity, not benchmarks.
Our FUP/QoS for networking works like this: some locations may have slightly different limits due to local differences, but in general, if you use 100 Mbit/s for 1 hour on average, we will globally limit your speed to 90 Mbit/s. If you then use 90 Mbit/s for the following hour on average, we will limit you to 80 Mbit/s, and we repeat this until you are down to 10 Mbit/s. For each hour you are 10 Mbit/s under the current cap, we add 10 Mbit/s back, until you are back to 100 Mbit/s. This helps to keep bandwidth available for everyone, while catering for genuine, suitable TierHive use cases.
If you need always-on, always-available, 24x7 high bandwidth over 100 Mbit/s, then TierHive is not the right option for you.
TierHive gives you a global meshed /24, and a subnet from 10.0.0.0/8, meaning if your VPS in France has the IP 10.254.200.12, and your VPS in Singapore has 10.254.200.13, they can access each other as if they were on the same LAN, and have full port-range access between locations with each other.
We don't allow it, except through known mail relays on port 587.
TierHive is a project run by Kernel Keepers LTD. It is not a giant cloud infrastructure run by corporate monsters, we try to do our best. If you are likely to lose any sleep or get stressed by occasional sub-par performance or downtime, it's probably best you pay a bit more and go with a host not trying to fill this basic, non-enterprise-grade market gap. While we do everything we can to give you a good experience, you need to acknowledge the limitations of the service. It is aimed at development, tinkering, homelab, hobby, and self-hosting, not always-on, perfect uptime. We have done what we can with free load balancers and services to make it possible for you to make your services redundant over multiple locations, so it can, in some cases, be more dependable than a single enterprise-grade VPS, but only if you do the work to make it that way.
On most occasions, we give you free credit to test the service before you ever have to pay a penny. We really appreciate people doing this, as it sets your expectations before you have to pay. However, if you use a disposable email address and a datacenter registered IP, we don't give you free credit, because sadly we got hit hard very early on by bot farming of free tokens.
We do not collect anything we don't genuinely need to provide the service. We even use a merchant-of-record system for payments via FastSpring, so technically you are never our direct customer, you are theirs, they just send us a signal that you have paid, and we add the tokens to your account. We do not want your data, we don't want to know who you are. We have no objection to you using a free or disposable email, or a VPN, etc., but we also have to protect ourselves, and these can be used to exploit our goodwill.