CloudFlare and CDN for cloud failover

Forums › IT Pro and developers › CloudFlare and CDN for cloud failover

1 | 2 | 3

timmmay

20574 posts

Uber Geek

Trusted

Lifetime subscriber

#1104173 7-Aug-2014 15:28

Ragnor: Why would you run it from home and failover to some form of cloud/hosting provider instead of just running it from the cloud/hosting provider in the first place?

Running from home just has too many single points of failure (power, internet, fire, theft etc).

Yes it costs money to host things properly but presumably you have a business model or plan in mind where when you have x customers it exceeds the y cost of running the servers. Most new online services run at a loss initially and require investment of time and money to build reputation and customers. Skimping on hosting will damage your reputation at some point most likely.

The whole reason Azure, S3 or even a small VPS at some hosting provider exist is to allow you to start small/cheap and scale up if you don't want to start with a high cost initial cost of owning your own servers and co-locating them.

OP already said it's because of cost - they need a bunch of CPU and RAM, which is expensive in cloud solutions like amazon. Yes lots of points of failure at home, but OP also said it fails rarely, and that's why they want to fail over to the cloud.

Everyone knows it's not an ideal setup, hosting in a data centre has many benefits, but it looks like that option's off the table.

jasnz

41 posts

Geek

#1104205 7-Aug-2014 15:53

@jnimmo Actually I called HD. The guy was very nice and knowledgable. As it turns out we couldn't actually get the 4U price since we have 4 X 1U servers. So it would actually be more like NZ$1000 using their web configuration. But he said very likely they could drop the price substantially with a combined hosting of 4 X 1U but they needed all the details. I have done this enough times that it probably isn't worth doing this exercise again. But I can say they seemed quite knowledgeable and friendly and their rates are inline or possibly better than others here in NZ. In any case it is likely to work out to be not so cheap. So I would recommend them by the sounds of it...

@Ragnor As commented further up the chain, unfortunately we have VM's which require the use of lot of memory. This is quite expensive under most cloud hosting arraignments. I guess as we monitor speeds, etc and find we can't keep up then we simply flip up to the cloud while we sort the problem out. The cloud prices are falling but things are still a lot more expensive there for now. Right now most of our VM's sit with very low utilisation but we still need them. Most cloud apps service a whole bunch of people and only need one VM. For instance lets say you were creating a ToDo service online like Wunderlist or something. Then you have a few machines or VM's which are used to near capacity and then you simply expand as you need and you are fully utilising each VM.

Instead think of say creating a Zimbra instance each one of which needs about 4 Gigs of memory. Then you have multiple small companies each which needs their own instance, but each instance is used only maybe 5 to 10% of the core it is on... Running this in the cloud is kind of hard to do... In fact if you look for hosted Zimbra it is kind of expensive for I think exactly this reason... So that is the explanation again.

Indeed for colocation we would instead probably buy a much more modern server which has 100+ gigs of memory in it and a good number of cores. Right now our servers are older and only have 32G's each.

@freitasm Interesting point! I'll have to check that. We do say we run a business from our home. We do have the power to the room on a serrate mains line, etc... But thanks for the reminder!

Ragnor

8218 posts

Uber Geek

Trusted

#1104223 7-Aug-2014 16:22

About HD, heard too many horror stories about them. The owner seems like a grade A "dishface" refer to previous HD (HostingDirect) related threads on Geekzone and GPforums if you want to know more.

There are plenty of other much more well established co-location providers eg:: ICONZ, Orcon, Vocus, The Data Centre, Datacom

1U is usually around $80-100+GST per month depending on what else is included (ip address, traffic etc), half rack is usually in the $500-600+GST range in my experience, probably cheaper if buying larger amounts.

timmmay

20574 posts

Uber Geek

Trusted

Lifetime subscriber

#1111098 19-Aug-2014 09:58

I'm doing some reading about AWS this week. It looks like Amazon's DNS system, Route 53, can potentially help here. It can do system health checks and route to both systems in the Amazon cloud and outside it. All I've read is the intro right now, but give me a few days and I may be able to help some more.

I'm a solution architect between contracts and could potentially help with this free of charge - it looks better on the CV to be doing something than nothing :)

jasnz

41 posts

Geek

#1111201 19-Aug-2014 11:55

@timmmay. We had thought about / considered route 53, but from the feedback and from other reading around it sounds more likely that we should use CloudFlare. But thank you for the great offer of help! I will private message you!

Cheers,
Jas

timmmay

20574 posts

Uber Geek

Trusted

Lifetime subscriber

#1151438 10-Oct-2014 16:59

Ok I've been doing a lot of AWS study recently, and I'll be sitting the architect exam pretty soon. Another solution occurred to me today.

In short, you could use Route53 as DNS, which will route to your primary servers. Route 53 supports health checks, basically it opens a URL on the web server and if it gets anything other than a HTTP 200 OK it can route all queries to a different server. It integrates with AWS so based on CloudWatch and auto scaling groups you could immediately ramp up your capacity while the health checks are saying the primary site isn't healthy, then it will turn the resources off once it's working again. Route 53 is incredibly cheap, you might pay a few dollars a month, pricing here.

You could also run in a "pilot light" configuration, where 1% of the traffic is serviced by AWS, making sure it's all working right all the time. You could potentially do that with a cheap T2.micro instance which is $10 a month - pricing here. That usage level is free for the first year.

I'm wondering if we could integrate with CloudFlare as well... possibly, though it wants to control DNS. That would take a bit more thought.

If you want me to think about this more send me a PM. I'd need to know what kind of severs you run (ie 3 web, 1 app, 1 database and the general technology), RTO and RPO (acceptable downtime and how recent the data has to be if you fail over). It'd be fun to put what I've learned into practice, happy to help no charge partly because it'd look great on my CV :)

1 | 2 | 3