Bulletproof Hosting
Discussion
Hi all, i'm running a group of websites at present and have been continuously let down by my hosting provider. My business requires 24/7 uptime, just an hour of downtime can cost me hundreds of pounds.
Can anyone recommend a decent hosting provider offering guaranteed 100% uptime, with decent bandwidth, php etc and not at a ridiculous cost?
Can anyone recommend a decent hosting provider offering guaranteed 100% uptime, with decent bandwidth, php etc and not at a ridiculous cost?
The one that sponsors this and the business forum might be a start.
Recommended by 32 PH'rs and rising daily
www.oilypages.com/recommendations.asp?id=5342
Independent Review Site
http://OcHostReview.co.uk
www.Register1.net/vds.php
Let me know if you have any queries.
J
Terms and Conditions said:
SERVICE AVAILABILITY
# We shall use reasonable endeavours to provide continuing availability of the Server and the Services but we shall not, in any event, be liable for Service interruptions or down time of the Server.
To be honest you're unlikely to find many hosts that go beyond that. Those that offer SLAs will usually allow an hour in any given month and if they exceed it start paying proportions of your monthly fee back. Everyone will allow themselves scheduled downtime too at quiet times of the day. Salesmen might like to push the idea of "100% uptime" but it just doesn't happen - too many variables, both in and out of the host company's control.
I'd go more on reputation. An expensive host promising the earth but failing to deliver and giving you back a proportion of your fees as compensation is rather less useful than one that makes no promises but just works a lot better.
>> Edited by sjg on Thursday 30th March 19:44
theres no such thing as 100% uptime. show me a company who guarantee it and I'll show you a liar.
if as you say you lose hundreds of pounds even if there is only 1 hour downtime, then i would suggest that you need to invest in a proper solution, rather than looking for cheap and cheerful hosting where the company promises the earth for peanuts.
a proper solution, need not cost the earth either. a load balancer and 3* 1U machines should do the trick and set you back about £5-6k. The running costs in something like Telehouse would be around £110 in space per month and say £200 per month for 2Mb of uncontended international bandwidth (approx 650GB per month) delivered over 100M or GigE.
if as you say you lose hundreds of pounds even if there is only 1 hour downtime, then i would suggest that you need to invest in a proper solution, rather than looking for cheap and cheerful hosting where the company promises the earth for peanuts.
a proper solution, need not cost the earth either. a load balancer and 3* 1U machines should do the trick and set you back about £5-6k. The running costs in something like Telehouse would be around £110 in space per month and say £200 per month for 2Mb of uncontended international bandwidth (approx 650GB per month) delivered over 100M or GigE.
JamieBeeston said:
I love your prices.. if only you could buy branded solutions for that.. and I expect the hosting prices to rise dramatically soon given Telecity buying out Redbus and jacking all prices 70-500%
I'll sell to you at that if you want. Sounds strange but for smaller customers we pass on the rack space at cost.
We never bothered with Redbus. Redbus was modelled after Telehouse North in Coriander Avenue, hardly suprising considering the history of various characters. Essentially, all the problems that Telehouse North has were copied at Redbus, whereas Telehouse actually learned some lessons from the North building and built the newer East building with a few key changes.
JamieBeeston said:
Recommended by 32 PH'rs and rising daily
www.oilypages.com/recommendations.asp?id=5342
Well if it's good enough for PH, it's good enough for me! Thanks to everyone for their comments. The key requirements are great up time 99.9%< and fast loading regardless of traffic, plus installation of some additional software etc. This service seems to cover all this under the VDS Pro package - is that what everyone would recommend?
MyNameIsMud said:
Recommended by 32 PH'rs and rising daily
www.oilypages.com/recommendations.asp?id=5342
Well if it's good enough for PH, it's good enough for me! Thanks to everyone for their comments. The key requirements are great up time 99.9%< and fast loading regardless of traffic, plus installation of some additional software etc. This service seems to cover all this under the VDS Pro package - is that what everyone would recommend?
7 Day money back Guarantee if you find its not to your liking... not had anyone take us up on it to date eheheh. Hundreds of PH'rs use Register1, which is why we're happy to give back in the form of Advertising / Sponsorship / Charity Prizes etc.
Do let me know if you have any queries.
J
sparkythecat said:
Hi all, i'm running a group of websites at present and have been continuously let down by my hosting provider. My business requires 24/7 uptime, just an hour of downtime can cost me hundreds of pounds.
Are you a porn king then?
Now there's a fun Job..
How does one apply?
tinman0 said:
theres no such thing as 100% uptime. show me a company who guarantee it and I'll show you a liar.
if as you say you lose hundreds of pounds even if there is only 1 hour downtime, then i would suggest that you need to invest in a proper solution, rather than looking for cheap and cheerful hosting where the company promises the earth for peanuts.
a proper solution, need not cost the earth either. a load balancer and 3* 1U machines should do the trick and set you back about £5-6k. The running costs in something like Telehouse would be around £110 in space per month and say £200 per month for 2Mb of uncontended international bandwidth (approx 650GB per month) delivered over 100M or GigE.
That is a bit strong tinman. there are planty of companies who will offer tht kind of SLA, but it wil come at a price bescause it is not easy to deliver.
100% uptime can't be achieved with 3 1U machines and a load balancer. What if the load balancer fails, how are you switching in the network because the load balancer won't be designed to do that (even if it is capable of it) then for a setup like that you'll need some protection so a decent hardware firewall would be needed and potentialy some Intrusion Detection. all this will need to be deployed in a fully redundant architecture with every device being paired in an Active passive configuration to be able to guarantee 100%. Power will need to be truly diversely delivered from different phases to both primary and secondary sets of hardware, servers will need to be good enough to support NIC teaming and have redundancy built in in terms of power supplies and cooling. You'll need to choose your data centre carefully because some have historicaly filled their space unevenly meanng that N+1 redundancy is compromised in some areas so power and cooling availability is not within building tolerances and a 100% guarantee could not be given.
Connectivity will need to be delivered through a fully fault tolerant network idealy with a properly multi homed peering setup to avoid carrier issues. International bandwidth is a bit of a misnomer, I know of at least one major global carrier whose peering distribution switches are CAT5513's with gig fibre uplinks without the HSRP enabled. That's huge contention in laymans terms even on Tier 1 bandwidth. Peering with a suppliers Access router is the only way of actually getting connectivity from the core of the network and it means a dedicated gig port on a Juniper (or similar) core router. The cards for that alone cost £10K then you ned to commit to at least 300Mbps of transit to be allowed a Gig port.
You can go further, but it means splitting or mirroring the site over 2 data centres and gloabl load balancing using BGP or something like a Cisco GSS.
Most people who claim they can do this sort of thing are claiming it because they know what they are talking about and because they have done it. it is not a cheap co-lo project and it is normally not necessary, but I take issue with the suggestion that people who offer it are liars and I suggest that a 'proper' solution is somewhat more complicated than just ploking a load balancer in front of more than one server.
dans said:Can you elaborate on that a bit for me?
...every device being paired in an Active passive configuration to be able to guarantee 100%.
I'd say that Active/Passive cannot give 100% uptime:
1. System is available
2. Active device fails
3. Failure is detected
4. Passive device assumes active role
5. System is available
Depending on the device and on the service being provided, steps 2,3,4 can be quick, so the outage can be short, but a short outage is not the same as 100% uptime.
What you're describing is a system with no single point of failure. This doesn't ensure 100% availability, but what it does is reduce the probability and likely duration of an outage. How far you have to reduce these before you can afford to commit to 100% availability is a matter of economics.
dans said:
That is a bit strong tinman. there are planty of companies who will offer tht kind of SLA, but it wil come at a price bescause it is not easy to deliver.
100% uptime can't be achieved with 3 1U machines and a load balancer. What if the load balancer fails, how are you switching in the network because the load balancer won't be designed to do that (even if it is capable of it) then for a setup like that you'll need some protection so a decent hardware firewall would be needed and potentialy some Intrusion Detection. all this will need to be deployed in a fully redundant architecture with every device being paired in an Active passive configuration to be able to guarantee 100%. Power will need to be truly diversely delivered from different phases to both primary and secondary sets of hardware, servers will need to be good enough to support NIC teaming and have redundancy built in in terms of power supplies and cooling. You'll need to choose your data centre carefully because some have historicaly filled their space unevenly meanng that N+1 redundancy is compromised in some areas so power and cooling availability is not within building tolerances and a 100% guarantee could not be given.
Connectivity will need to be delivered through a fully fault tolerant network idealy with a properly multi homed peering setup to avoid carrier issues. International bandwidth is a bit of a misnomer, I know of at least one major global carrier whose peering distribution switches are CAT5513's with gig fibre uplinks without the HSRP enabled. That's huge contention in laymans terms even on Tier 1 bandwidth. Peering with a suppliers Access router is the only way of actually getting connectivity from the core of the network and it means a dedicated gig port on a Juniper (or similar) core router. The cards for that alone cost £10K then you ned to commit to at least 300Mbps of transit to be allowed a Gig port.
You can go further, but it means splitting or mirroring the site over 2 data centres and gloabl load balancing using BGP or something like a Cisco GSS.
Most people who claim they can do this sort of thing are claiming it because they know what they are talking about and because they have done it. it is not a cheap co-lo project and it is normally not necessary, but I take issue with the suggestion that people who offer it are liars and I suggest that a 'proper' solution is somewhat more complicated than just ploking a load balancer in front of more than one server.
My example is very simple example of doing a high uptime setup. Of course the point of failure becomes the load balancer, but then theres nothing stopping you adding a failover and so forth. but then the costs start mounting up.
Sorry, but anyone who claims 100% is a liar. They hide behind a little clause that says they only guarantee these things that are under their control but cannot claim 100% on events outside their control, eg a DDoS, or a bgp routing problem. I know of at least one company who offered 100% uptime. Superb claim, and the day it went wrong wasn't because the network failed - it was because they forgot to renew the customers domain.
tinman0 said:
Sorry, but anyone who claims 100% is a liar. They hide behind a little clause that says they only guarantee these things that are under their control but cannot claim 100% on events outside their control, eg a DDoS, or a bgp routing problem. I know of at least one company who offered 100% uptime. Superb claim, and the day it went wrong wasn't because the network failed - it was because they forgot to renew the customers domain.
I think you may be misinterpreting the '100% availability' claim. You don't need to able to certain of achieving 100% availability in order to guarantee it. All the guarantee does is make you liable if an outage occurs. In fact I don't think you could ever be certain of achieving 100% availability; there is always the possibility of an outage if enough things goes wrong. Think of it like a guarantee on a washing machine. The makers who offer a guarantee aren't saying that none of their machines will ever go wrong, they're just saying they are willing to pay if it does.
GreenV8S said:
Sorry, but anyone who claims 100% is a liar. They hide behind a little clause that says they only guarantee these things that are under their control but cannot claim 100% on events outside their control, eg a DDoS, or a bgp routing problem. I know of at least one company who offered 100% uptime. Superb claim, and the day it went wrong wasn't because the network failed - it was because they forgot to renew the customers domain.
I think you may be misinterpreting the '100% availability' claim. You don't need to able to certain of achieving 100% availability in order to guarantee it. All the guarantee does is make you liable if an outage occurs. In fact I don't think you could ever be certain of achieving 100% availability; there is always the possibility of an outage if enough things goes wrong. Think of it like a guarantee on a washing machine. The makers who offer a guarantee aren't saying that none of their machines will ever go wrong, they're just saying they are willing to pay if it does.
Indeed, and anyone who is serious about their 100% or even Five nines will have an extremely large penalty clause built into the SLA that will make the providers Wince when they pay up (very punitive, not just x days service free)
The client will pay through the nose for such an SLA though!
It is not possible to provide 100% indefinitely (though it obviously is over a given period!) as even with fully redundant kit and BGP you're still looking at upto a 3 min window for the route to stop being broadcasted / relayed.. sure some people on the alternate feed(s) will still be able to get to you, but not everyone.
This is as close to 100% as you can get.
JamieBeeston said:
Indeed, and anyone who is serious about their 100% or even Five nines will have an extremely large penalty clause built into the SLA that will make the providers Wince when they pay up (very punitive, not just x days service free)
At Hughes we had a support team of half a dozen engineers. Two outages of half an hour would have wiped out our annual support fees. And after that it started to get *really* expensive ...
jamieboy said:
...every device being paired in an Active passive configuration to be able to guarantee 100%.
I'd say that Active/Passive cannot give 100% uptime:
1. System is available
2. Active device fails
3. Failure is detected
4. Passive device assumes active role
5. System is available
Depending on the device and on the service being provided, steps 2,3,4 can be quick, so the outage can be short, but a short outage is not the same as 100% uptime.
Fair point - I ought not to sound off with a hangover. In practice the failover should be pretty much transparent if the setup is properly configured and the users of most systems will not notice the changeover. There are few circumstances in which deeper reliability is needed and you can build active/active, but that is a pig.
Gassing Station | Computers, Gadgets & Stuff | Top of Page | What's New | My Stuff