1

I've been bitten by this too many times. You find a well known hosting provider, read comparative reviews (and pass by on any company that has influence in those review), they promise 99.99% uptime, you pay a reasonable price (not the cheapest) and after you bought the server or VPS, they end up delivering less than 95% uptime. I'm not naming any names, but these are some well-known companies.

How can you independantly verify the validity of their uptime claims before handing over the money?

Martijn
  • 356
  • 5
  • 13

2 Answers2

3

There's a difference between the track record in the past and using that as the prediction for the future. You can use independent parties like netcraft that allow you to get an idea how long sites have been up.

Something to also pay attention to is whether they give you a "best effort" estimation or an SLA. If you get an SLA they need to show you how they measure the availability and give a clear statement on the penalties that apply when they don't make the goal (e.g. a percentage of the sum you paid).

In general if you are talking about individual machines (both physical or virtual) I would not expect to get availability better than 99.5%: you will need downtime for maintenance etc.

Then get your hands on multiple smaller machines (VPS) in different availability zones and combine them into something stronger.

Theuni
  • 958
  • 5
  • 15
  • +1 for investigating the SLA and penalty clauses for failure to deliver. A company that has a strong service level agreement (and will pay substantial penalties if they fail to deliver) is likely to be quite reliable, as they obviously don't want to be giving you money. (The flip side is they're also likely to be quite *expensive* as ensuring a high level of availability requires lots of costly redundancy.) – voretaq7 Dec 08 '12 at 04:01
2

You can't... There are some hints and steps you can take, though...

  • Ask about the hardware manufacturers/types in use. Brand names == good. Supermicro, well... depends on the implementation. :)
  • Ask about the common redundancies. At the server-level, power, fans, disk. At the facility level, power, cooling, network, generator, etc.
  • Ask the provider to give a current customer reference contact or two...
  • Ask about past service outages and how they've handled things.
  • Lower your expectations... if this is a small low-cost VPS, the incentive isn't there for the provider to bend-over backwards to guarantee uptime. At which point, you can build redundancy into your application or just deal with the occasional outage. Depends on how much you're paying...
ewwhite
  • 197,159
  • 92
  • 443
  • 809
  • Maybe I'm missing something, but I've only ever heard good things about Supermicro... – EEAA Dec 07 '12 at 18:06
  • @EEAA I think the Supermicro depends on the implementation. I find it to be lacking, coming from an HP background. But it can be made to run as well... – ewwhite Dec 07 '12 at 18:09
  • Someone that offers 99,99% SLA for a VPS (in a reasonable price) is obviously lying. – FINESEC Dec 07 '12 at 18:25
  • @FINESEC - whether or not it's a VPS has **nothing** to do with its reliability. Reliability has everything to do with the underlying infrastructure and implementation. That said, over the lifetime of one of my Linode servers, it's in the neighborhood of six nines over the span of 7 years. That's pretty darn good for $20/month. – EEAA Dec 07 '12 at 18:39
  • Whoever you go with, get the guarantees in writing, especially their definition of terms they use ("uptime" is different for different people), how they arrive at their promised numbers, and how they will be measured/reported on. Also, agree on a recourse (financial penalty, etc) if terms of the contract are not met. – KM. Dec 07 '12 at 18:53
  • Can't really agree with that because statistically VPS is more prone to hardware errors than a dedicated server. Your VPS might have 99.9999% availability but that says very little about average availability. Linode offers 99.9% uptime for a reason ;-) – FINESEC Dec 07 '12 at 19:03