Is there any method that ensures 100% or 99,9999% uptime?

Except the promises of the various hosts?;)

I think dreaming is the only solution for 100% uptime.
100% uptime is impossible, you will have sometimes some kind of downtime... May be you can keep 100% for weeks as lots of hosts can like ours ;) but in years you will have some time some kind of downtime.

I think the only way somebody could offer a 100% uptime is if they did the following:

Dual back up power
Clutering
Hot swap adapter cards, drives
RAID 5 on all arrays

And lots of $$$$$

Almost forgot, several different upstream providers --

Yep, that is probally almost impossible unless you charges about $1,000 per month per client ;)

Greetings:

Uptime is a factor of several issues:

* Network

* Hardware

* Operating System

* Applications

* Management procedures

Unless all of the above have 100% uptime, then you are looking at the least common demonitor to determine the actual uptime (management impacts in terms of maintenance et all).

Also, there are some providers that guarantee their network 100% uptime. This is next to impossible and more marketing hype.

Thank you.

Originally posted by dynamicnet
Also, there are some providers that guarantee their network 100% uptime. This is next to impossible and more marketing hype.


With my limited knowledge/understanding on how things really work, I would like to give some of my 2 cents.

It's not really impossible nor it's meant for a marketing hype. InterNAP offers us 100% uptime guarantee and after looking at their model, I do think it's actually possible. To start with, they have fully redundant core routers in which each of their carriers come into one of these routers in a fully redundant way. So Core Router A will have UUNet1, ATT1, etc and Core Router B will have UUNet2, ATT2, etc (this explanation is a oversimplification, btw). UUNet1 and UUNet2 (as well as ATT1 and ATT2, etc) come to InterNAP through fully diverse path (say, UUNet1 from east and UUNet2 from west). Of course to ensure the fully working of their routers/equipments, they too have fully redundant power grid and power generators with UPS backup to always kick in once there is a power outage.

And for their customers, they give two 100 mbps drops each cabinet, one from each of their redundant routers. With HSRP (Hot Spare Router Protocol) of these two redundant ethernet drops from two redundant routers, the customer can practically achieve 100% network uptime without any problem, well ... hopefully.

Is this just a marketing hype? I don't think so? Why would they invest such a big $$$ and go into all these problems just to have a marketing hype?

again, just my 2 cents.

cheers,
:beer:

Greetings:

Yes, there are those that offer it. But you don't get it.

Our enterprise network monitoring system is used by a number of the large providers. No one to date has 100% up time.

Even the ones who state redudant this and redudant that usually have one focal point that is a single point of failure.

Thank you.

Not even Yahoo have been able to reach 100% uptime. That said, I think if you mirror redundant servers through different upstream providers on completely separate networks, and can guarantee power for several days even if there is an outage, you should be almost there. Just be prepared to pay a lot of $$$ -- you won't get that for 10 bucks. ;)

That does not mean it's just merely a marketing hype. No ... Think of uptime guarantee as a warranty/insurance policy. When InterNAP (or UUNet, Level3, etc) offers 100% uptime, it's not meant to be a marketing hype. They have the infratructure that capable of delivering this uptime. So when they can not deliver 100% uptime as they guaranteed, they have to refund customers and for big cutomers like yahoo, they will have to refund big money too, I suppose.

Marketing hype will be like this: A host offering unlimited transfer/bandwith although the host knows very well that he has no power/resource/ability to deliver unlimited transfer/bandwith. In this case the word "unlimited' is used to trick people to signup with the host. But when a customer utilizes 2GB of transfer/month, the host needs to suspend/terminate this customer because it can't actually deliver 2GB of tranfser not to mention unlimited transfer. This is, IMHO, a marketing hype.

But to say UUNet, Level3, or InterNAP offering a 100% uptime guarantee as a marketing hype is an understatement and a bad analogy. They strive very hard and invest a huge amount of money to deliver their uptime guarantee, don't compare this with the marketing hype of those "unlimited bandwith hosts" that do not even have the ability to deliver what they offer.

cheers,
:beer:

Just to chip in, we get 100% network and power from our data centre, and they have delivered it impeccably since we installed 10 months ago - hard to beat :)

At the end of the day, you get what youy pay for - the idea that 100% uptime is difficult and costly will get few arguments, but the fact that it is possible, and backed-up by SLA's is a fact of modern day hosting,

Best regards,
Steve

They claim they have had 100% uptime for failsafe customers, it says right on their main page:
www.conxion.com
Conxion has provided 2280 Continuous 100% Uptime Days, Since Feb 24, 1996 for ALL Conxion's FailSafe Customers - With No Exceptions!

I dont know much about them, I found them when getting a file from download.com

Many very large enterprises download's seem to origionate from conxion so I assume they are pretty good.

Does anybody know anything about them?

--Simon

I know they host (or used to) most of Microsofts downloads...

Steve

Greetings:

Let me state that when one promises 100% it does not mean it is delivered.

I base this on 7 years experience along side our network monitoring system which monitors several of the majors.

I mean marketing hype from the stand point they are making a promise on the basis they know they are going to be paying out credits against it from non-delivery.

Furthermore, network uptime is only one component out of several in terms of being able to advertise what uptime you provide or a hosting company provides.

Thank you.

What if the network on which sits your network monitoring system fails?

IN fairness though, we've seen it delivered for almost a year, continuous uninterrupted service on both network and power. Of course periodic scheduled maintenance/installations take individual servers down for a few minutes every couple of months, but that is covered by the SLA too :)

Steve

Greetings:

We have mutliple monitoring stations.

So if you want to believe in the next to impossible, be my guest.

Thank you.

>So if you want to believe in the next to impossible, be my guest.

LOL :)

All I'm believing in is what I have personally experienced from my provider for the last 10 months. Nothing more, nothing less.

It's hard to argue that it's impossible when its a fact :)

Steve

Originally posted by Satj2000
100% uptime is impossible.

Not true. Conxion have had 100% uptime for 5 years and they guarantee uptime of 99.999%.

Greetings Steve:

"All I'm believing in is what I have personally experienced from my provider for the last 10 months. Nothing more, nothing less."

What are you using to validate what you have experienced?

Our parent company hosted an on-line bank for three years.

They left to work with a more local company.

The CEO talked with the CEO of the parent company personally and stated there was not a service issue; Mark (the CEO of the bank) stated they had no down time in three years.

His experience was pure perception as there was quantifiable and identifiable down time during those three years.

Mark, just didn't experience it or see it for himself.

Therefore, unless you have a system in place which verifies your experience, it is more perception than verified reality (no offense intended).

Thank you.

We have external monitoring, plus logs, data from the machines themselves, plus, more importantly, customer perception.

If all our customers, and our staff, who all monitor the services could not detect any downtime, I tend to agree that there was none :)

Steve

100% uptime?

Not that difficult surerly?

Buy good equiptment, pay good wages, make your staff happy, and ask them how its done.

Oh, use at least 3 different backbone connections with 3 way neworks on each. Loads of servers, mirrors etc.

You would have thought banks would have been 100%, but im sure I heard somewhere that they are in fact not, they leave a portion of downtime for essencial maintanance or core upgrades. I may be wrong, but it does make a lot of sense, but they do put that aside so they are hoping for 100% of scheduled uptime?

Anyway, creating self supporting networks is not hard if you have the money. All it takes is advanced network knowledge, experience and the will to go and find the docs on the net.

Internal networks does not pose a problem as this can be overcome by many different systems, having multipule connections to backbones and redundacy for them is the key.

I know its possible.
:)

Greetings:

Next to impossible does not mean impossible. However, it does mean highly unlikely.

The hardware has to have 100% uptime.

To a degree you have raid and redudant power supplies. But unless you get into special motherboards, you can have faulty processors or memory and other components.

And you can bypass that with load balancing and clustering. All of this adds up.

The operating system and applications have to offer 100% uptime.

This means get rid of your standard distributions. You now have to use specialty versions. Yes, they are out there.

Then your administration team has to have all of the procedures and policies in place so that when a piece of the rendundant pictures goes bad, they can replace it, document it, etc.

100% uptime? Possible. But costly.

100% uptime in the context of your standard shared hosting provider? Next to impossible.

Rackspace, among the very best, offers 99.999% on their network yet deliver slighly less.

Most shared providers don't utilize load balancing, speciality hardware, etc.

And as one person shared in the thread dealing with how often you reboot your servers, that in two years time (if dealing with Linux), you will be rebooting several times (which means at least once during that year if not several) just to update the kernel.

In ending, I do believe there are people (and they have shared it) that believe they get 100% uptime. My comment to them is that belief and reality are separate. You have to have proof of the experience for it to go beyond perception.

Thank you.

So,

It is possible, so possible that a lot of networks actualy can deliver it, and not just claim to. Furthermore by delivering it they get good customers who pay enough for these operations to maintain and fund research into operations further.