Web Hosting Talk : Web Hosting Main Forums : Cloud Hosting : Incorporating Cloud Computing as part of your business model

[leggomygreggo]

If you wanted to offer cloud computing services to your customers would you simply get a data center and interconnect a bunch of servers together along with a NAS drive and sell based on the customer need?

Sorry for the newbie question but I think this would be a great thing to offer to customers..just kind of unsure about how the whole thing works

[TheProxyHoster]

if you have the clientele for it then it is a good idea. It takes a very large initial investment to get it up and running

[leggomygreggo]

I figured it would be a large investment but is the idea right? Could anyone explain how a company would go about setting up a cloud? I dont really plan on doing something like that anytime soon...just curious

[01globalnet]

You can offer cloud solutions without re-inventing the wheel - you can resell from companies like Amazon, Mosso, Cartika, TheGridlayer etc. etc.

[mattle]

Point of curiosity: Is anyone using cloud storage for their web hosting clients currently? I've considered it as you can get more space at a cheaper rate than renting a server with large drives. I'm just not sure if the network-mounted drive space will be fast enough for serving up web hosts and VPS's.

[hexahost]

I would say Using Cloud for Backup is a good idea, But its Not a good idea to serve web directly from cloud hosting.

[leggomygreggo]

Why would you say that using cloud hosting to serve web directly isnt a good idea? Just wondering why it might be a bad idea

[mattle]

FYI, I finally got some white papers from ThePlanet listing their timing data (and some of their competitors...but probably to be taken with a grain of salt...)

(all figures in MB/sec)
The Planet (hosted server to local cloud node): Up-36.92, Down-38.01
Amazon (EC2 server to S3 storage): Up-11.93, Down-10.13
Mosso (Cloud server to cloud files): Up-4.03, Down-1.77

For the sake of comparison, here's some independent testing data I noticed on some of my machines (using hdparm)

(buffered disk reads)
15k SCSI : 60.43
10k SCSI : 84.54
EIDE (7200RPM) : 58.34
EIDE (Unknown) : 29.81

So, as far as speed is concerned, good cloud storage is slightly faster than a crappy IDE drive. I'd like to see how these tests would fare on an NFS mount in the same rack...

[blessen]

Originally Posted by leggomygreggo

If you wanted to offer cloud computing services to your customers would you simply get a data center and interconnect a bunch of servers together along with a NAS drive and sell based on the customer need?

Sorry for the newbie question but I think this would be a great thing to offer to customers..just kind of unsure about how the whole thing works

If there is market and requirement for Cloud computing then you can go for it. Business originate because there is a requirement.

I would advice to use services from providers like Amazon,Mosso etc. These are the common ones and the most talked about ones in this forum.

[maknet]

I tend to agree with most people here.

I wouldn't go into cloud computing unless my clients had a need. If i get one inquiry, i'd probably refer do a reseller thing or a make-shift solution depending on need.

But if there really are a lot of people then i would consider investing in the infrastructure.


Lawrence

[yangyang2036]

I don't think cloud hosting has anything substantially special or better than traditional hosting. Having it would be cool but without it it's no big deal at all.

From a business perspective, any other moves than reducing costs should be thought through seriously. I suppose you are just thinking cloud hosting is the right idea because it's cool and cutting edge rather than it brings lower costs to you or better performance to your customers?

[plumsauce]

Originally Posted by mattle

FYI, I finally got some white papers from ThePlanet listing their timing data (and some of their competitors...but probably to be taken with a grain of salt...)

(all figures in MB/sec)
The Planet (hosted server to local cloud node): Up-36.92, Down-38.01
Amazon (EC2 server to S3 storage): Up-11.93, Down-10.13
Mosso (Cloud server to cloud files): Up-4.03, Down-1.77

For the sake of comparison, here's some independent testing data I noticed on some of my machines (using hdparm)

(buffered disk reads)
15k SCSI : 60.43
10k SCSI : 84.54
EIDE (7200RPM) : 58.34
EIDE (Unknown) : 29.81

So, as far as speed is concerned, good cloud storage is slightly faster than a crappy IDE drive. I'd like to see how these tests would fare on an NFS mount in the same rack...

Those figures don't even tell half the story. There is latency to consider. How fast can you start serving a 1kb css file from the time the request hits the frontend?

[Nex7]

Latency wasn't part of the tests done, true. However, I can tell you informally that the figures are similar, relationally, to the speed tests; for much the same reason the speed tests came out like they do.

The Planet to cloud test is the most 'closely tied' sets of infrastructure, physically speaking (any server at The Planet is likely to be less than 2-3 network hops from the local node of the Storage Cloud, all within The Planet's internal network). Amazon S3 to EC2 is the second closest, likely staying on Amazon's internal network but having more physical disparity, and Cloud Server to Cloud Files is the farthest (Cloud Server, in case you don't know, is rebranded Slicehost that Rackspace bought; which isn't even in the same state in the US as the storage of Cloud Files, as I understand it).

So network latency is likely to be equivalent to the speed tests in terms of who wins and by how much.

That said, network latency is only half the story there, as well; there's also application latency to consider (The Planet Storage Cloud, Mosso Cloud Files and Amazon S3 ARE all application-based storage engines.. a computer is answering your requests, not a hard drive directly). If you want to know which is best, I'd suggest testing!