TierHive - Update post - 8th September 2026

AnthonySmith

A bit of an update post, as it has been a while, I accidentally posted this on the end of an old offer. I can't find the old TierHive update post, I think I accidentally transformed an offer and got them all moxed up, and it was long anyway, so I'm starting a fresh one.

HAProxy

Upgrades were done to the HAProxy environment; some mistakes were also made, and we took out a chunk of the HAProxy mesh for a while but we quickly had the others take over, so no one seemed to notice a little bit of added latency for a while.

We improved a few things on HAProxy:

• log rotation was tweaked as they handle way more traffic than expected initially and disks were filling up.
• We started explicitly adding domain:80 and domain:443 as well as domain to the ACL's along with a tunnel timeout, which helps people running websockets behind our HAProxy
• We had to implement tracker detection in HAProxy, as sme customers thought "No torrent activity" meant it is ok to run trackers hitting us with 10k requests p/second; this traffic is now auto-forwarded to GCHQ.
• Added a little more RAM and compute, and we had a few nodes not auto acceting http 2 defaulting to 1.1, that has been fixed.

Singapore

Singapore is pretty overloaded again; more capacity ordered and upgraded this time to cope with the demand and also adding storage (HDD) options.

I know it's not doing well right now in terms of performance. For some reason, we started getting a massive influx of new users in the region. We usually get around 30 - 50 new customers per day on average; for some reason, in the past week, that has jumped to 250 - 300 per day and Singapore was hit hardest by demand. (Nice problem to have, but still a problem)

Stable base, BETA and API

We are now starting to have the first conversations needed to start exposing the API to some degree, but to be clear, it still won't be generally available until we go to BETA, we have agreed that the BETA date will set when we have had 30 days of stable + 30 days.

What this means is that once we have gone 30 days without an issue being reported or found that turns out to be something non-trivial on our side that we need to fix, we will internally set the 30-day countdown to move into BETA. We will announce this if it happens, and assuming nothing crazy happens in the following 30 days that shakes our confidence, we will move to BETA and start fully releasing the API.

I feel like I have been giving everyone an arbitrary date of early 2027; it's nothing but a best guess, and it's not really right. We move to beta when we feel stable enough to be in BETA, and we will likely stay in BETA for a number of years or as long as is needed.

Once in Beta, we will naturally take on more responsibility and accountability for issues and failures and will start to put together a basic SLA. We also plan to get our own ASN registered at this point and start looking at better IPv6 coverage, we will also start looking to gradually replace the hardware over time, moving to EPYC-based platforms as standard offering more RAM and Cores.

New Locations & More RAM

Just in case anyone is not aware TierHive is now in:

• New York, NY (USA)
• Kansas City, MO (USA)
• Vint Hill, VA (USA)
• Hillsboro, OR (USA)
• Beauharnois (Canada) (Up to 8GB Ram)

• Frankfurt (Germany) (Up to 8GB Ram)
• London (UK)
• Warsaw (Poland)
• Amsterdam (Netherlands)
• Paris (France)
• Roubaix (France)
• Gravelines (France) (Up to 8GB Ram)

• Singapore (SG)
• Tokyo (Japan)
• Sydney (Australia) (Up to 8GB Ram)

Static Hosting

Not a lot has changed; we did increase the stability a lot, and we don't get as many issues reported these days, it just kind of works; around 400 users are on the static hosting platform actively, which is fun to see. I will probably release a special offer when that hits 1000.

General Numbers

We stopped counting "Live users" unless they had a VPS live as the volume of trial users is obviously high so the metric has changed a bit, TierHive has created around 15,000 virtual servers and has 3209 Live service accounts currently (A user with at least 1 active VPS) and around 12,000 live accounts (Has added credit and has created a virtual server within the last 90 days)

We still prune users after 90 days of inactivity if they have only had free/trial credit; this is always ongoing; this may also be something we stop doing in BETA.

What's really good to see is that there is a clear pattern of people using it to build virtual servers on demand, short-lived, via recipes and their own startup recipes, only paying for the time they need and then powering off to reduce compute for us and costs for themselves, or just outright deleting and redeploying on a clear pattern (weekends, etc)

We find that Fridays and Sundays have at least 25% more compute time used than other days.

Now & Next

We are struggling a little bit with demand and scale so a lot of the work being done now is to optimise rather than new features and this will probably stay the focus until we move into beta.

My personal goal is to have 24 locations by the end of 2026; it's not feeling likely now, but I do have a target list:

• Vancouver (CA)
• Toronto (CA)
• Chicago (US)
• Miami (US)
• Dallas (US)
• Denver (US)
• Los Angeles (US)
• Ashburn (US)
• Zurich (CH)
• Hong Kong (HK)
• Stockholm (SE)
• Oslo (NO)
• Seoul (KR)

No specific order; I will just pick off that list based on timing and availability. These are locations where prices remain fairly stable; we have already identified the upstream host for the location.

So that's what is up in the land of TierHive.