Cloud hosting rocks

Cloud hosting is really cool. You should check out Amazon Web Services. Pay for what you use, bring nodes up and down as you please for storage, computing, etc. It is pretty awesome actually.
 
From a corporate stand point; cloud hosting is cool, but you have to be aware of the dangers. You are essentially handing off some pretty critical infrastructure tasks to a 3rd party ( maintenance and backups ), as well as allowing your data to mingle "out there". You have no way to verify that no one else has access to it, because it's not local. Many people have physical access to the hardware, and thus your data.

My company is researching the pros and cons of cloud hosting right now which is why I'm posting this.
 
Indeed. There's also legal concerns. If I have a server in my own building, then I have a pretty good expectation of privacy over the server and the data on it. If I don't even own the server and the data is on machines managed by someone else in the cloud, then legally it is hard for me to defend against searches and seizures which end up releasing that data.

I don't think the hand off issue can be stressed enough. If you're running your own servers, you can drive to the data center and figure out exactly what's wrong when they go down. If you're in the cloud, then you've got to call their support, get in line with hundreds of other customers, and talk to a first-line support person who's going to start out by asking you if you're using the right password.

The potential for data loss in sub-cotnracted storage is very real, and directly impacts your customers in a significant way. It's really their data after all, right? Meanwhile, you're not in direct control of that experience.

I'm sure there are applications where it's viable and useful, but the risks are so great that it's difficult for me to understand when some of those situations might be.
 
I agree with mikeblas. Our servers hold a lot of private health information including SSNs. There is no way we will ever trust that information to a 3rd party source.
But even if we did, that whole MS Sidekick debacle was a huge blow to my confidence in placing all my important data in the cloud. As a very last resort backup, sure, but not before I have my own set of backups under my direct control.
 
For 'unimportant' sites like blogs and whatnot, cloud-based hosting is a hoot. You get the cost savings of shared hosting and the ability to handle massive traffic spikes (Slashdot/Reddit/Digg).
 
I agree with mikeblas. Our servers hold a lot of private health information including SSNs. There is no way we will ever trust that information to a 3rd party source.
But even if we did, that whole MS Sidekick debacle was a huge blow to my confidence in placing all my important data in the cloud. As a very last resort backup, sure, but not before I have my own set of backups under my direct control.
Honestly, I view cloud computing in the same light as I view data lines from the phone company; I won't use them in a serious environment unless I have an SLA detailing my rights and expectations ( along with the all important repercussions if they aren't met ).

I would also trust critical data to a service i trusted ( along with the afore mentioned SLA ). Think: online backup services. There are just a couple considerations that one needs to have.
 
An SLA isn't a guarantee of service. It tells you what happens when service isn't available. If you look through the Amazon site, the SLA provided is incredibly weak. They say you'll get 99.9% uptime. There are 43,200 minutes in a week, so each month you might be down for 43 minutes.

If you're down for 42 minutes, it doesn't mean a thing. If you're down for 43 minutes, I don't think it means anything. If you're down for 44 minutes, the difference is a minute and you're credited for your pro-rated loss, so it's still won't matter if that's not more than a dollar. Since they're only giving you 10% of the month (unless they're more than 1% down), you really don't have anything to hold against your business losses or your customer complaints. Maybe your credit will be enouhg to by a candy bar, or if you're a really big user, a box of copier paper.

I don't see anything that makes a cloud service any more reliable than shared hosting, and that includes sudden load situations. The cloud would just back-end your queries, so your shared web host is still under load by answering queries, firing queries off the the cloud back-end, and waiting for responses. This is actually more latency, right? If you want to avoid that latency, you're obliged to cache things on the front-end server, and you'll burn resources on that, too.
 
I wish it was cheaper, $100/mo is incredibly expensive compared to shared hosting still.

I would move myself and a few clients over to it if it was not so expensive.
 
An SLA isn't a guarantee of service. It tells you what happens when service isn't available. If you look through the Amazon site, the SLA provided is incredibly weak. They say you'll get 99.9% uptime. There are 43,200 minutes in a week, so each month you might be down for 43 minutes.
That's assuming a lot however; my SLAs are 5 9s. I wouldn't commit anything to the cloud without it, nor would I commit anything to it if I didn't have a full company tour ( mainly tech tour, but I'll take everything ).

Of course, the work would have to fit the application. I'm not one to shoehorn a project into a solution if it doesn't fit, just because of a buzz word ( a fact which has gotten me yelled at by more CxOs than I can remember ).
 
That's assuming a lot however;
What's assuming a lot? I simply stated facts--no assumptions. I mean, sure: I'm assuming there are 30 days in a month, and 24 hours in a day. But ...
 
What's assuming a lot? I simply stated facts--no assumptions. I mean, sure: I'm assuming there are 30 days in a month, and 24 hours in a day. But ...
You are assuming that's what I'd accept, 3 9s of up time.
 
I have no preconceived notions about what you'd accept.

Maybe it's your point that you could negotiate a better SLA with Amazon, or find a provider who can do better. Maybe that's true ,but I don't think it would happen for the casual blogger -- or even a pretty serious site customer. Once you're paying a few hundred bucks per month, you have to figure out if that money isn't better spent on leasing your own hardware and managing your customer's experience directly, yourself.
 
I think the point of the original poster was that is a great service for a pretty good price. Not every company can afford to build and staff it's own data center. Yes, I know that having your own data center would be ideal, but there is a whole "can of worms" with that setup too, not to mention the major expense (hardware, software,redundant power, redundant connections to internet,...)

It is not too different a situation as using a Dedicated Hosting provider, except the hosts are virtualized and can scale pretty quickly. Granted and there could be risks with information leakage between VM's, but that is a much smaller threat than just the general threat of having any "server" connected to the internet and everything that goes with that.

With that said Dedicated VM's (Cloud Computing) may not be applicable for every case, but I think for any situation that would normally call for physical dedicated servers, the virtual server may be a step better just from a flexibility and cost standpoint.
 
Indeed, there are very few companies that build and staff their own data centers. Most use co-location services, or rack-oriented hosting services. No staff is necessary, really, beyond the IT guys; and all the major expenses you mention are provided by the point-of-presence provider.
 
I came to read why the OP thought it was so awesome. Too bad was interested in reading it in the remaining few minutes I have at work. Anyway, I can't even begin to consider it for a client my company works for. I'm working on software clients and two database solutions for a client under strict data access restrictions. Imagine that, designing a database for data I can't sample.
 
I probably should have elaborated on why I thought it was awesome. Anywho, I agree with those of you who wouldn't consider it because of the lack of infrastructure control. I believe that cloud hosting is awesome for an application where shared/managed/dedicated server hosting would be the alternative.

Por ejemplo: Rackspace Cloud VS. 1and1 Managed Server

* Cost: basic cloud processing, 2gb transfer per day = way cheaper than 1and1 managed server. Winner: cloud

* Backup control: I have a daily, weekly, and on demand backup copy of my sites with the cloud server. When 1and1's managed server crashed, their backup was corrupted and lost an entire site. Winner: cloud

*Downtime: the $200/month 1and1 server goes down multiple times a week for hours. Sometimes days. Rackspace updates this blog with problems and resolutions. Check out some of the response times. Winner: cloud.

*Scalability: Cloud - go from a server with 256MB of RAM, 1/64th of a quad core proc to a server with 15872 mb of RAM and a full quad core proc within a minute. Awesome. Winner: cloud

*Ease of use: The traditional server wins this one, as the cloud is just a blank Linux install with no cPanel or preconfiguration. Installing CentOS/Apache/PHP/mySQL/mail doesnt take long though. Winner: traditional host


Thats why I think my new cloud server is awesome.

Of course, my own datacenter full of these, guarded by laser beam sharks would be superior to the cloud server.
 
I probably should have elaborated on why I thought it was awesome. Anywho, I agree with those of you who wouldn't consider it because of the lack of infrastructure control. I believe that cloud hosting is awesome for an application where shared/managed/dedicated server hosting would be the alternative.

Por ejemplo: Rackspace Cloud VS. 1and1 Managed Server

* Cost: basic cloud processing, 2gb transfer per day = way cheaper than 1and1 managed server. Winner: cloud

* Backup control: I have a daily, weekly, and on demand backup copy of my sites with the cloud server. When 1and1's managed server crashed, their backup was corrupted and lost an entire site. Winner: cloud

*Downtime: the $200/month 1and1 server goes down multiple times a week for hours. Sometimes days. Rackspace updates this blog with problems and resolutions. Check out some of the response times. Winner: cloud.

*Scalability: Cloud - go from a server with 256MB of RAM, 1/64th of a quad core proc to a server with 15872 mb of RAM and a full quad core proc within a minute. Awesome. Winner: cloud

*Ease of use: The traditional server wins this one, as the cloud is just a blank Linux install with no cPanel or preconfiguration. Installing CentOS/Apache/PHP/mySQL/mail doesnt take long though. Winner: traditional host


Thats why I think my new cloud server is awesome.

Of course, my own datacenter full of these, guarded by laser beam sharks would be superior to the cloud server
The points and past history presented only seems as a compelling argument to get off of 1and1's hosting service, and not necessarily be replaced with a Cloud implementation. With those "wins" for Cloud hosting, what driving factors or client business trends make them significant or appropriate?

In a related note, I've really enjoyed the debate and perspectives presented so far in this thread.
 
The "wins" for cloud hosting over shared hosting in backup recovery and uptime are completely anecdotal and not endemic to the technologies.
 
The "wins" for cloud hosting over shared hosting in backup recovery and uptime are completely anecdotal and not endemic to the technologies.
Agreed. I was more interested in the rationale that went into the comparisons ;)
 
Have you guys seen http://vps.net/

Virtual private cloud server £15/mo

I'm currently considering reselling this hosting to bring in some extra money on the side of the web development i currently do.
 
Looks a little iffy. How do they make overclocked 7.2 GHz servers run reliably? I think in their case, "cloud" is just a marketing word and doesn't speak to the architecture.
 
I'm guessing that these servers they have 2 quad core cpu's in each say it was 3.2ghz per core theoretically each server would be 25.6ghz.
 
Nah, that can't be it; clock rate is neither theoretically nor practically additive.
 
I'd like to bump these since I'm doing a small amount of research on this for a project in school.

I'm curious if anyone has had any updates to their findings presented earlier in this thread.
 
I don't think it sucks any less than it did six months ago.
 
In regards to Rackspace, do you know if their Cloud Servers product supports for setting up a n-tier load-balanced environment using one of their own CServers as the load balancer?

So...

1 CServer RHEL/CentOS Linux running something like HAProxy1.4
3 CServers RHEL/CentOS Linux running Apache2.2
1 CServer RHEL/CentOS Linux MySQL5

Is there a solution for multiple MySQL boxes in either LB or mirror?

No SAN/NAS or shared storage baked into this solution.
 
On the note of the 7.2Ghz... I think just like Terremark, in this solution you "carve" out VM's up-to the CPU processing you've purchased--With probably with a single-instance limit.
 
In regards to Rackspace, do you know if their Cloud Servers product supports for setting up a n-tier load-balanced environment using one of their own CServers as the load balancer?

So...

1 CServer RHEL/CentOS Linux running something like HAProxy1.4
3 CServers RHEL/CentOS Linux running Apache2.2
1 CServer RHEL/CentOS Linux MySQL5

Is there a solution for multiple MySQL boxes in either LB or mirror?

No SAN/NAS or shared storage baked into this solution.

The Rackspace cloud hosting is more abstract than that. You have your PHP app hosted there & it talks to their DB and their cloud system handles how many servers need to be involved. It's sort of like Google AppEngine - the system dynamically allocates resources as needed. If you want to actually have virtual machines, rather than dynamically allocated compute resources, you need to go with somebody more along the lines of Amazon.
 
Amoeba, you may be referring to their CloudSites product, not CloudServers. With CloudServers, you create a machine with lets say 2cpus and 8gb of RAM and that's what you get. Unlike CloudSites where your code is running across a vast farm of LB servers.

The Rackspace cloud hosting is more abstract than that. You have your PHP app hosted there & it talks to their DB and their cloud system handles how many servers need to be involved. It's sort of like Google AppEngine - the system dynamically allocates resources as needed. If you want to actually have virtual machines, rather than dynamically allocated compute resources, you need to go with somebody more along the lines of Amazon.
 
There's the additional problem of true data manipulation. We do some extensive reporting off of our data, and are having headaches because not every hosted solution provides you with an ODBC connection or a clear-cut API through which to connect to the information you need.
 
Amoeba, you may be referring to their CloudSites product, not CloudServers. With CloudServers, you create a machine with lets say 2cpus and 8gb of RAM and that's what you get. Unlike CloudSites where your code is running across a vast farm of LB servers.

Oh. I guess they've expanded their offerings since I last looked.
 
For $10 a month my Debian VPS is 50gb storage, 1tb transfer, 1gb ram. I have a small sql database and some scripts that do aggregation on it... nothing major but it's been fine.

Why would I move to cloud if it's more expensive?
 
Back
Top