I am currently working on starting a full service managed server provider with own racespace, switches, etc., I plan to aim for rackspace's standards one day. For those of you that look for a managed server, what kind of features/services do you find the most appealing?
Keep in mind this will not be a budget host.Extremely fast, reliable network. Not just good US routes, but also great routes to Europe and Asia. N+1 redundancy at least with hot failover (VRRP / HSRP) edge/core routers. Would be good to use bandwidth providers with a history of uninterrupted connectivity, such as Internap and Mzima. Good DDOS protection wouldn't hurt either.Extremely fast, reliable network. Not just good US routes, but also great routes to Europe and Asia. N+1 redundancy at least with hot failover (VRRP / HSRP) edge/core routers. Would be good to use bandwidth providers with a history of uninterrupted connectivity, such as Internap and Mzima. Good DDOS protection wouldn't hurt either.
I think this pretty much sums it up right there.I am currently working on starting a full service managed server provider with own racespace, switches, etc., I plan to aim for rackspace's standards one day. For those of you that look for a managed server, what kind of features/services do you find the most appealing?
Keep in mind this will not be a budget host.
A strong understanding of complex setups, something that rackspace does not offer. Being able to say "I want 10 servers. I want Four mysql servers in an active-active replication scenario. I want two different bandwidth mixes, one that is low latency and high quality for my website, and one that is cheap for static media. I want load balancing between these two front-end servers. I'd also like the option of doing my own routing and getting a full BGP feed from you".
This is something that Rackspace won't do, but an intelligent managed services provider should be able to do for a price.Could I get some input from people that are looking or have looked? I need ideas of services to offer.The type of thing I'd want to see is more in the backend side of it.A fully-featured monitoring backend for clients to take a peek at and view their servers and statistics in an easy to use panel.Beyond that the previous folks in the thread summed it up well!Could I get some input from people that are looking or have looked? I need ideas of services to offer.Monitoring and proactive response.Fully managed. At least os and major services, like php, mysql, control panel on linux.Hardware/bandwidth/management price options.And don't forget the 100% uptime sla. If you want to offer premium servicess, you must offer 100% uptime sla.And don't forget the 100% uptime sla. If you want to offer premium servicess, you must offer 100% uptime sla.
A 100% uptime SLA is worth the paper it's written on. The rule of thumb is for any infrastructure, multiply your costs by 10 for every 9 you add. 100% over what? Business hours? A month? A year?
The 100% uptime is something marketing people provide to customers who don't know any better, sadly.A 100% uptime SLA is worth the paper it's written on. The rule of thumb is for any infrastructure, multiply your costs by 10 for every 9 you add. 100% over what? Business hours? A month? A year?The 100% uptime is something marketing people provide to customers who don't know any better, sadly.Internap offers a 100% SLA on their IP. They live up to it, too.Nothing wrong with SLA'ng a network to 100%, but I do agree with you if you were referring to hardware...(even clustered setups).this is both an easy and a hard questionon one side its down to v. good monitoring (both network and host based - ie disks, cpu temp etc), good communication (somethings going wrong, or looks to be going wrong, so you tell the customer in advance), and being able to offer the widest range of solutions to customers. on the flip side its very difficult to articulate to a customer what they want, and persuade them that your solution is what they needthey're paying a premium, so they need to get bang of their buck - redundancy, response and proactive support - you'll find that a lot of the customers will need their hands held, and you'll need to draw the line about what they can and cant do with 'their' machine. its no longer a machine they're hiring, but a service and this will need to be communicated to them - they come to you and say ive got an ecommerce site, with a pile of databases, and i dont want it to go down, and i want it to work quickly, and you say to them you can do that with 4 machines - 2 front end and 2 back end, and then give them the alternatives, and why your solution is best..just some musings/thoughts from my experiences really - its definitely not about technology - thats just a small part of the overall pictureInternap offers a 100% SLA on their IP. They live up to it, too.
I seem to recall a rather long outage in their Seattle PNAP maybe 6 months ago that took out LiveJournal for 2-3 days.
<!-- m --><a class="postlink" href="http://www.eweek.com/article2/0,1895,1751832,00.aspI">http://www.eweek.com/article2/0,1895,1751832,00.aspI</a><!-- m --> seem to recall a rather long outage in their Seattle PNAP maybe 6 months ago that took out LiveJournal for 2-3 days.
<!-- m --><a class="postlink" href="http://www.eweek.com/article2/0,1895,1751832,00.asp">http://www.eweek.com/article2/0,1895,1751832,00.asp</a><!-- m -->
What about network outages not influenced by power outages? Those have never happened AFAIK.Even akamai can't provide 100% uptime, but 100% uptime should be the target. Not 99.999999 uptime, but 100% uptime. It's a good marketing strategy and aiming to really offer 100% uptime will force you to provide the best services you can.What about network outages not influenced by power outages? Those have never happened AFAIK.
I'm sure I've read somewhere about Internap PNAP outages before - it's not impossible. As has been said previously, 100% uptime SLA's are nothing more than marketing gimmicks (i.e. we don't meet the SLA, we credit your account up to 100% of your monthly fee, depending on the length of the outage) - but that doesn't cover for the fact that the site/server was still down for that ammount of time and that there is revenue lost.
You've proven that point yourself in a past post about a Rackspace outage in LON2. They offer 100% uptime SLA.Even akamai can't provide 100% uptime, but 100% uptime should be the target. Not 99.999999 uptime, but 100% uptime. It's a good marketing strategy and aiming to really offer 100% uptime will force you to provide the best services you can.
Sure - but what happens if something outside of your direct control affects the network?You've proven that point yourself in a past post about a Rackspace outage in LON2. They offer 100% uptime SLA.But they had an SPOF (one fiber provider) the whole time.If anything I proved that rackspace is lying that they are fully redundant.Let's not take this thread off topic.Time and time again, Internap and Mzima have 100% uptime at their POPs or damn close to it, that's all I'm trying to say.Time and time again, Internap and Mzima have 100% uptime at their POPs or damn close to it, that's all I'm trying to say.
Thank you for proving my point for me: "...or damn close to it" - so even they are not 100% uptime by your own admission. I can't think of one provider that has had true 100% uptime for an expansive period of time.Bandwidth suggestions besides mzima/internap? catering to asia?BTN/Tiscali/DTAG/NTT?BTN/Tiscali/DTAG/NTT?
Is NTT really all that great out there?NTT is a carrier that has it's roots in Asia. Mzima also has many good asian routes.We use Internap for our managed customers in San Francisco, and they are by *far* our favorite ISP.. However, we have had about 30 minutes of outage on their network in the past 2 years. That's an awesome uptime, but it is by no means a 100% uptime.
As ISPs go, Internap is the best I've ever dealt with, actually. What's more important isn't the marketing smoke & mirrors of a 100% uptime but communications. Internap lets me know :
- If there are errors on my port
- If I am experiencing an out of pattern bandwidth spike
- If they are planning maintenance
- If they are undergoing unplanned maintenance
- What above maintenance is for
- The impact of maintenance
- The post-mortem of the maintenance
When I call up Internap's NOC, I get a real router engineer.. Not a $10/hour support monkey, and not an answering machine... That's one of the niceties that puts Internap head and shoulders above all the other ISPs I've used.Bandwidth suggestions besides mzima/internap? catering to asia?Mzima already has some great Asian routes, they peer with many Asian carriers through PAIX (in Palo Alto). But, if you want even better routes PCCW Global (aka BTN) is good (especially for the price they charge). DTAG and Tiscali are not that great in CA (I assume Los Angeles is where you are looking?) but awesome in the Northeast (Washington, DC - New York).Honestly you do not need any more carriers than Mzima & Internap. Though Internap isn't really a carrier anyhow, instead they give you access to 8 tier-1s (including NTT - who monopolizes the Asian bandwidth market for the most part).We use Internap for our managed customers in San Francisco, and they are by *far* our favorite ISP.. However, we have had about 30 minutes of outage on their network in the past 2 years. That's an awesome uptime, but it is by no means a 100% uptime.
Are you including maintenance in there by any chance?Layer0 - I believe 20-22 minutes of that downtime was due to maintenance. Again, an SLA is merely marketing fluff.
Keep in mind this will not be a budget host.Extremely fast, reliable network. Not just good US routes, but also great routes to Europe and Asia. N+1 redundancy at least with hot failover (VRRP / HSRP) edge/core routers. Would be good to use bandwidth providers with a history of uninterrupted connectivity, such as Internap and Mzima. Good DDOS protection wouldn't hurt either.Extremely fast, reliable network. Not just good US routes, but also great routes to Europe and Asia. N+1 redundancy at least with hot failover (VRRP / HSRP) edge/core routers. Would be good to use bandwidth providers with a history of uninterrupted connectivity, such as Internap and Mzima. Good DDOS protection wouldn't hurt either.
I think this pretty much sums it up right there.I am currently working on starting a full service managed server provider with own racespace, switches, etc., I plan to aim for rackspace's standards one day. For those of you that look for a managed server, what kind of features/services do you find the most appealing?
Keep in mind this will not be a budget host.
A strong understanding of complex setups, something that rackspace does not offer. Being able to say "I want 10 servers. I want Four mysql servers in an active-active replication scenario. I want two different bandwidth mixes, one that is low latency and high quality for my website, and one that is cheap for static media. I want load balancing between these two front-end servers. I'd also like the option of doing my own routing and getting a full BGP feed from you".
This is something that Rackspace won't do, but an intelligent managed services provider should be able to do for a price.Could I get some input from people that are looking or have looked? I need ideas of services to offer.The type of thing I'd want to see is more in the backend side of it.A fully-featured monitoring backend for clients to take a peek at and view their servers and statistics in an easy to use panel.Beyond that the previous folks in the thread summed it up well!Could I get some input from people that are looking or have looked? I need ideas of services to offer.Monitoring and proactive response.Fully managed. At least os and major services, like php, mysql, control panel on linux.Hardware/bandwidth/management price options.And don't forget the 100% uptime sla. If you want to offer premium servicess, you must offer 100% uptime sla.And don't forget the 100% uptime sla. If you want to offer premium servicess, you must offer 100% uptime sla.
A 100% uptime SLA is worth the paper it's written on. The rule of thumb is for any infrastructure, multiply your costs by 10 for every 9 you add. 100% over what? Business hours? A month? A year?
The 100% uptime is something marketing people provide to customers who don't know any better, sadly.A 100% uptime SLA is worth the paper it's written on. The rule of thumb is for any infrastructure, multiply your costs by 10 for every 9 you add. 100% over what? Business hours? A month? A year?The 100% uptime is something marketing people provide to customers who don't know any better, sadly.Internap offers a 100% SLA on their IP. They live up to it, too.Nothing wrong with SLA'ng a network to 100%, but I do agree with you if you were referring to hardware...(even clustered setups).this is both an easy and a hard questionon one side its down to v. good monitoring (both network and host based - ie disks, cpu temp etc), good communication (somethings going wrong, or looks to be going wrong, so you tell the customer in advance), and being able to offer the widest range of solutions to customers. on the flip side its very difficult to articulate to a customer what they want, and persuade them that your solution is what they needthey're paying a premium, so they need to get bang of their buck - redundancy, response and proactive support - you'll find that a lot of the customers will need their hands held, and you'll need to draw the line about what they can and cant do with 'their' machine. its no longer a machine they're hiring, but a service and this will need to be communicated to them - they come to you and say ive got an ecommerce site, with a pile of databases, and i dont want it to go down, and i want it to work quickly, and you say to them you can do that with 4 machines - 2 front end and 2 back end, and then give them the alternatives, and why your solution is best..just some musings/thoughts from my experiences really - its definitely not about technology - thats just a small part of the overall pictureInternap offers a 100% SLA on their IP. They live up to it, too.
I seem to recall a rather long outage in their Seattle PNAP maybe 6 months ago that took out LiveJournal for 2-3 days.
<!-- m --><a class="postlink" href="http://www.eweek.com/article2/0,1895,1751832,00.aspI">http://www.eweek.com/article2/0,1895,1751832,00.aspI</a><!-- m --> seem to recall a rather long outage in their Seattle PNAP maybe 6 months ago that took out LiveJournal for 2-3 days.
<!-- m --><a class="postlink" href="http://www.eweek.com/article2/0,1895,1751832,00.asp">http://www.eweek.com/article2/0,1895,1751832,00.asp</a><!-- m -->
What about network outages not influenced by power outages? Those have never happened AFAIK.Even akamai can't provide 100% uptime, but 100% uptime should be the target. Not 99.999999 uptime, but 100% uptime. It's a good marketing strategy and aiming to really offer 100% uptime will force you to provide the best services you can.What about network outages not influenced by power outages? Those have never happened AFAIK.
I'm sure I've read somewhere about Internap PNAP outages before - it's not impossible. As has been said previously, 100% uptime SLA's are nothing more than marketing gimmicks (i.e. we don't meet the SLA, we credit your account up to 100% of your monthly fee, depending on the length of the outage) - but that doesn't cover for the fact that the site/server was still down for that ammount of time and that there is revenue lost.
You've proven that point yourself in a past post about a Rackspace outage in LON2. They offer 100% uptime SLA.Even akamai can't provide 100% uptime, but 100% uptime should be the target. Not 99.999999 uptime, but 100% uptime. It's a good marketing strategy and aiming to really offer 100% uptime will force you to provide the best services you can.
Sure - but what happens if something outside of your direct control affects the network?You've proven that point yourself in a past post about a Rackspace outage in LON2. They offer 100% uptime SLA.But they had an SPOF (one fiber provider) the whole time.If anything I proved that rackspace is lying that they are fully redundant.Let's not take this thread off topic.Time and time again, Internap and Mzima have 100% uptime at their POPs or damn close to it, that's all I'm trying to say.Time and time again, Internap and Mzima have 100% uptime at their POPs or damn close to it, that's all I'm trying to say.
Thank you for proving my point for me: "...or damn close to it" - so even they are not 100% uptime by your own admission. I can't think of one provider that has had true 100% uptime for an expansive period of time.Bandwidth suggestions besides mzima/internap? catering to asia?BTN/Tiscali/DTAG/NTT?BTN/Tiscali/DTAG/NTT?
Is NTT really all that great out there?NTT is a carrier that has it's roots in Asia. Mzima also has many good asian routes.We use Internap for our managed customers in San Francisco, and they are by *far* our favorite ISP.. However, we have had about 30 minutes of outage on their network in the past 2 years. That's an awesome uptime, but it is by no means a 100% uptime.
As ISPs go, Internap is the best I've ever dealt with, actually. What's more important isn't the marketing smoke & mirrors of a 100% uptime but communications. Internap lets me know :
- If there are errors on my port
- If I am experiencing an out of pattern bandwidth spike
- If they are planning maintenance
- If they are undergoing unplanned maintenance
- What above maintenance is for
- The impact of maintenance
- The post-mortem of the maintenance
When I call up Internap's NOC, I get a real router engineer.. Not a $10/hour support monkey, and not an answering machine... That's one of the niceties that puts Internap head and shoulders above all the other ISPs I've used.Bandwidth suggestions besides mzima/internap? catering to asia?Mzima already has some great Asian routes, they peer with many Asian carriers through PAIX (in Palo Alto). But, if you want even better routes PCCW Global (aka BTN) is good (especially for the price they charge). DTAG and Tiscali are not that great in CA (I assume Los Angeles is where you are looking?) but awesome in the Northeast (Washington, DC - New York).Honestly you do not need any more carriers than Mzima & Internap. Though Internap isn't really a carrier anyhow, instead they give you access to 8 tier-1s (including NTT - who monopolizes the Asian bandwidth market for the most part).We use Internap for our managed customers in San Francisco, and they are by *far* our favorite ISP.. However, we have had about 30 minutes of outage on their network in the past 2 years. That's an awesome uptime, but it is by no means a 100% uptime.
Are you including maintenance in there by any chance?Layer0 - I believe 20-22 minutes of that downtime was due to maintenance. Again, an SLA is merely marketing fluff.