New on LowEndTalk? Please Register and read our Community Rules.
All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.
All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.
How can a host sell 1x dedicated core for $6.99 on a E3v5 system and make profit?
Did just see this thread https://www.lowendtalk.com/discussion/105576/1-dedicated-core-kvm-free-upgrade-to-ssd-and-e3-1240v5-6-99-month from @MSPNick there they are selling a hosting plan with 1 dedicated core of a E3-1240v5 system. I'm sure that actually is a 1 dedicated thread, and since E3 i a single CPU system, they can max have 8 customers per node with this deal. 8 x $6.99 = $55.92/mo
How is it possible to make profit on a "new" E3-1240v5 system when you not getting more then $56/mo?
For if they oversell and put more then 8 customers on the server, they are not getting a dedicated core, but a shared core.
Help me understand this if you can.
Comments
When you put 8 customet it's already shared core, if they owned the hardware maybe it's loss leader to their other service
Loss Leader/ Long break-even.
Just roughly throwing numbers out there: If a 1U colo works out to be $20/mo for the host (if they are running their own DC or have large #'s of racks, it can probably work out even lower) they would be netting ~$35/mo ($55-20). You can probably pick up a E3-1240v5 system for under $1000. Since there's no SSDs in the offer, and it doesn't look like they're offering RAID either, I would say $700-800 is a good price off ebay for something similar. If you buy for $800, then at $35 net/mo, you will break even in 23 months. Whether this is a good business plan or not, I won't comment... but if you're willing to wait long enough for ROI, pricing like that is certainly possible.
edit: forgot to mention shifty marketing tactics, but I'd like to hope this isn't the case with MSP.
check the advert: 1 Core @ 2.27GHz E3-1240v5
2.27GHz
Those things turbo up to 3.9 without breaking a sweat so that's 3.9 *8 /2.27 = 14 (ish)
14 * 6.99 = 97.86
But yeah, that's not really a dedicated 'core' or even a whole thread, seems a bit like CC/HVH marketing.
Buy one, if you see any %st which is perfectly normal for 99% of multi tenant environments but not when you have a dedicated/pinned core.
Edit: never mind, just checked the site, they are being sold as "Fair share CPU usage" so calling is 'dedicated' is just not true.
...Or they just scam you
...Maybe they dont have their own hardware -> maybe they get an offer like 39$/mo for a E3 and 16gigs of ram and 2x1tb hdd, they just use 8 THREADS, not CORES. for 6,99 which in conclusion is 8x6,99=55,92.
I think they just lie about 'dedicated cores' honestly.
They state it is a free upgrade. So I would guess it is provided at full speed after the upgrade. Need a confirmation from @MSPNick though.
maybe yeah, would make sense, what does not however is:
What are dedicated core KVMs?
Exactly what it say it is, you get dedicated core's for your applications
followed by: "Fair share CPU usage"
source: https://www.myserverplanet.com/dedicated-core-kvm.php
@AnthonySmith It's seems like their not upgraded plan is from a E5 system due to the CPU speed. But if you uograde you get a E3-1240v5 CPU, so then they can put only 8 customers on that plan.
@jonchun they actually offer SSD with the 1240v5 plan. But no info about RAID.
They £10.00 plan have two dedicated E3 cores.
How they can make the profit?
scam
I don't think dedicated core has to mean that it is dedicated to you.
Just put as many people on the node that nobody cares if any customer maxes a core/thread.
For an E3 I would probably put 20 or so people on there.
This story does my thinking about unlimited cloud storage. The most people won't use more 1TB but a few (<1%) use more or a lot more. But the avenge space of a user is still less then 1TB.
I think that they doe the same here. The most people won't use the core 100% for a long time.
but if the most people go use there core 100% and the avenge speed sucks. Then we go call it a scam and ask all or money back. I am waiting for the topic on LET.
So most likely there is lot more then 8 customers on a server like that?
See my edit and yes very likely if they don't want to lose money.
Anyone's guess.
Given that you get 2gb ram I'd say they put 32 customers on a node, given that a e3v5 supports up to 64gb ram.
It can be even more so with OVZ.
Keep in mind that you can be dedicated your own QEMU core, and limited to a percentage of timeslices, so your "dedicated core" is approximately at 2.2Ghz, while the host is 3.2Ghz+. OVZ also allows for stuffing more people on the system as it's tie direct into the kernel, reducing RAM and other overhead by comparison to KVM or "full" virtualization.
i hope that @MSPNick can share more information with us.
Microclouds can fit 8-24 E3s in 3U and save a ton on pricing too. Your 1U pricing for an E3 is way higher compared to the 1/8th of 1U for a 24 node Microcloud.
In most countries on can/must keep hardware 2 - 3 years on the books. After that one can sell them very cheaply or just keep it running at next to 0 pro-forma cost. Considering that most hw, particularly at the cheaper end of the market, actually runs for 5 (and not rarely even more) years, the cost distributes quite differently.
Assuming a period of years, one of it thrown for "free" to compensate for the occasional box that doesn't survive or that needs some repair, leaves us with 4 years or 48 months. That leaves us with about 20$/mo for hardware.
Next throw in bandwidth which is bought at about 2$/Tb and keeping in mind that 90% of customer do not even use a considerable fraction of their bw, they can sell lots of VPS w/2TB bandwidth where only 1 TB or less is actually used.
Finally, colo. A price of 1000$/rack incl. 100TB is not uncommon, which leaves us with 25$/unit incl. 2,5TB traffic.
So per box: 20$ hw, 8$ traffic, 25$ colo (power assumed included as those boxes consume little) -> 53$ / box / month.
Problem is that unless they are already a well going company in the market, at least the box must be payed front-up which brings in opex (I leave out capex as it's quite tightly coupled to income in that business).
In other words: Either they accept a loss for tax reasons or they cheat ("dedicated core") or they have bought cheap crap hardware. I personally guess it's a mix with the last point playing a prominent role. Hint: Upgrading the cpu and throwing in SSD upgrades usually has quite practical reasons, namely crappy old boxes that are pimped up with a modern cpu to something acceptable; and the SSDs are there for a reason, too: Power consumption plus marketing value.
(power, which is often not considered but plays a major role in cost)
That worm smells. I wouldn't bite.
I thought we were done with this breakup/makeup stuff!
You again, you evil bitch. You know perfectly well that your beautiful hair drives me crazy.
As for makeup. I don't use any. I'm nature-ugly.
TL;DR; dedicated vcore does not equal a (single) core on a dedicated.
in marketing terms it more likely means "dedicated to you" = we don't care if you use 24/7 whatever you get out of it (but no one said it's gonna be as much performance as a real single core might give)
Every beauty needs his beast..
Get a hotel room already and make sure to film everything, so that I can watch it later.
Reruns of Rerun so What's Happening?
Not in the UK, just not going to happen.
When mostly idle they will run at low power but when your selling servers specifically designed for people to run the cores 24x7 then that is the sort of customer you will get and your looking at the top end of power use possible for these boxes.
@AnthonySmith
Sure. But then 40 * 100W or so is within what's often included at mediocre colos.
The problem is also psychological. Bandwidth is something that people understand in terms of "the more you need the more you pay". Power, however, for 99,5% of clients is something that's just included. So, as a provider you want to find the sweet spot and hey, 80W (cpu TDP) is 25% less than 120W.
Plus it's a density problem. Many colos can hardly deliver even 10 KW per rack or 250W/RU. Now think of a dense blade (10U) (let alone micro cloud) system with 16 blades in 10 RU. With 10KW / Rack a cpu eating 80W is feasible, one with 2 * 130 isn't.
Which btw. makes me expect lots of Xeon D based systems/blades.
The other point you bring up is interesting. I personally think that still most customers won't go to the limit but I might be wrong (and you right).
My personal experience is that the vast majority of (even low end take whatever you can get at type) clients hardly ever really use the cpu and bandwidth they think they need.
Also, most (incl. many providers) do not fully understand the necessity of balance. I see a lot of waste in all areas.
As for uk colo pricing I might indeed be far off; I just don't know it well as I'm almost only looking at FRA/AMS/PAR for professional colos/dedis.
On bulk maybe it could work, why you would bother I don't know, but 1U colo with 0.5 maybe 1A at a push you are going to be paying £30 minimum in my experience and that is bargain basement, so about $40
In the US. In EU you pay power per A or per kWH and one A at 220V runs you 10-100EUR.
4kW power cost you easily 800EUR in Germany alone. Before VAT. This is not Ohio or Nevada.
Blades - not really, blade center buyers usually have no power problems and the Xeon-D are way behind E5.
Systems - yea, Hetzner and OVH have them, they are not bad and do 128GB RAM instead of 32GB on an E3 (pre v5).
I have yet to see anywhere that offers 1A for less than £49, however, if one was to own the hardware they could simply be pricing it out over 4-5 years