Ask HN: $20 per month dedicated server?
The problem (haven't verified) is that when you are measuring performance metrics, they can be hell lot misleading because of the extra s/w sitting between your app and h/w.
I think there is a need for a dedicated server hosting for startups which is priced around $30 per month (2-4 core CPU ~3Ghz, 1-2 GB Ram, 500G hdd, 1TB bandwidth). What do you think about this idea?
I haven't done the cost calculation but would like inputs from someone experienced. As a startup, will hosting on a dedicated server for 30$ per month be a exciting option? Would you take it?
Edit: ----- Since $39 option is already there, then let me drop the price further $20. Obviously I can drop because its still an idea but I really want to know the price point at which startups will be more inclined to rent a dedicated server instead of a VPS because obviously dedicated is better. let me know if I am missing something here.
10 comments
[ 2.4 ms ] story [ 40.3 ms ] threadhttp://www.ovh.com/us/dedicated-servers/
And 1TB per month is just 3mbps x 30 days, so that too can be misleading (1mbps x 30days = 330GB). They should tell you how many mbps you can use, not how many bytes you can ship.
TL;DR: The hardware is cheap. You _must_ have spare servers and plan for failure because they _will_ fail from time to time. Avoid dealing with their support people (just replace the server when there's some problem with it) and you'll be fine.
In more detail:
- Renew servers monthly. This will let you drop any server at any point and only lose the remainder of that month's worth of server cost.
- The hardware you get is cheap. This is generally not much of a problem, but you _must_ expect things to fail at some point.
- They handle most of these failures pretty quickly. A failed motherboard gets replaced in less than 30min. usually.
- The real PITA are the disk drives. They are _not_ enterprise disks, which means that their failure rate and performance loss over time is larger than what you are probably used to.
- It is very hard to convince OVH to replace underperforming but operational disks. Basically, it is hard to convince them about any degradation that doesn't halt the server. The time you must spend to convince them is just not worth it: replace the server and call it a day.
- Hence, you should always be ready to switch from one server to another. It's much easier to just rent a new server and cancel the older one than to make them acknowledge non-obvious hardware problems.
- A good strategy is to rotate servers every 1 or 2 years (tops) to avoid getting to this point.
- From time to time, a new server you rent is faulty. _Always_ test the hardware _extensively_ before you start using it.
- There will be outages from time to time, but they are rare. When this happens, some of your servers will go offline for as much as 1h. Most of these problems are caused by power supply problems, which means that your server will get shut down uncleanly and may have problems coming back up.
- Just for the fun: most of the outages are caused by them fking up while testing electrogen groups and/or backup power supplies. Sometimes we wish for them to not have any redundancy systems to test...
All in all, the service is very good for that price, but you _must_ be prepared to handle failures.
This is true regardless. For example, if your I/O benchmark is basically sequential, and your application's access pattern is basically random, then your benchmark is useless.
Ultimately, the only test that matters is whether a particular piece of hardware runs your particular application. So as a startup, you write your website, put it on a VPS, if it's slow then you get a bigger VPS or switch providers. The common case is that a VPS is good enough, and once you grow beyond it, that means you're getting enough revenue to afford an upgrade to a bigger VPS or dedicated server out of money coming in.
If you can't afford the upgrade when you need it, then it means either you're inadequately capitalized (you don't have enough investment to survive the bootstrapping regime where necessary server spending outstrips revenue), or your code is too inefficient for your business model to be viable (making $X in revenue costs more than $X worth of server resources, the "bootstrapping" regime has no end).
> 1-2 GB Ram
Upgrade this. RAM is cheap, and you get a lot of performance out of it.
> I haven't done the cost calculation
The available information at my fingertips says buying a full rack (40U), power included, costs $650 per month [1]. So that's $16.25 per 1U. That leaves $3.75 per month, or $180 total over a 4-year lifetime. Let's say you use a Raspberry Pi as the server, it costs $35. That leaves $145, or $3.02 per server per month, to pay for bandwidth, labor, billing/accounting, storage media, power cables, networking gear, and profit for the owner.
$20 is really aggressive pricing. $30 might be more reasonable.
If you have sunk costs, then you might be able to make it work. E.g. if you just signed a multiyear deal for a ton of rack space, but the project you were planning on using it for isn't going to fly due to your cofounder getting hit by a bus, you can regard the rent on server space as a sunk cost since you'll be paying for it anyway. Filling it with $20/month servers would almost certainly be better than letting it sit idle (though maybe not better than filling it with more expensive servers, subletting it to someone else, or having your company declare bankruptcy).
Can I say for sure one thing that between dedicated vs VPS you will surely choose dedicated?
Are there any downsides to dedicated servers from customer standpoint. I know from providers standpoint it's requires more work.
If they have the same specs and the same price, dedicated is probably better.
> Are there any downsides to dedicated servers from customer standpoint.
1. Price.
2. For a dedicated server I want a serial console and out-of-band rebooter, so I can fix it myself when I bork the bootloader or hang the system. This can all be done in software with VM's, but may require additional hardware for dedicated servers.
3. Monitoring of disk failure. On a VPS, the provider typically handles RAID, and is pro-active about replacing failed disks. On a dedicated server, RAID is up to the customer to manage, depending on your setup it might be the customer's responsibility to alert support that a disk is bad, and diagnosing a failed disk remotely might be challenging.