I've taken on a client with a news website that serves both static and video (http, not streamed) content, plus a couple of newsletters that are sent out regularly. After a few rounds of briefs and comments, I had specced out a server on Singlehop.com with the following characteristics:

.- Core2Quad @ 2.83Ghz
.- 6.00GB RAM
.- 1TB Hard disk
.- 5 Fixed IPs
.- 10TB Monthly transfer

, at a total cost of $314.00 monthly.

Some "consultant" then recommended my client towards Amazon EC2 saying its cheaper and better than a dedicated server. The Amazon price calculator (using average transfer rates) for a comparable system gives me a total of $626.92 per month if I use EC2 storage (risky...) and $634.42 if I use S3 storage.

However, if I prototype an entire year (using historical data), the Singlehop server is $3,768.00 per year (just 314x12) - but the Amazon server is a whopping $8,324.16, more than twice as much.

So am I missing something here or are we just below the economy of scale necessary to actually see the cost savings?

He must have said to migrate to EC2 because its based on cloud computing. While the other one is a dedicated server.

I will advise not to go with Amazon EC2 because of a lot of reasons which do not appear in the first shot. On a amazon server you can never get 2 IPs allocated to a single server.

Amazon is costly and a lot of people have lost data on their servers.

How can I get more than one IP address?

An Amazon EC2 instance is limited to only one public IP address for external use from outside the Amazon EC2 network. Each instance is also assigned a single IP address for internal use within the cloud.

You'll find more information about instance addressing in our tech docs.


--------------------- SEE AMAZON FAQ -------------------------

Can I manage my DNS within Amazon EC2?

All instances come with an internal and external DNS name. Amazon EC2 does not provide access to modify these DNS settings. If you would like to map an existing domain name to an Amazon EC2 instance you will need to use one of the many DNS management services that are available on the Internet today.

Within Amazon EC2, DNS requests for the external DNS name of an instance will resolve to the internal IP address of the corresponding instance.

When using your own domain name, we recommended mapping to the instance's external DNS name using a CNAME, not by using an A record pointing at the instance's IP address.


Does Amazon EC2 support reverse DNS lookups for email sent from my instance?

If you modified the DNS for your domain (mydomain.com) to include a CNAME to the external DNS name of an instance, and send an email to someone from your instance [user@mydomain.com] a reverse DNS lookup will result in the external EC2 DNS name, not mydomain.com.

###############################################################3

Cloud computing is high availability, and you can scale and copy your instances. It is the main advantage in comparison to any other hosting.

If you want we can also help you find a good cloud computing service which is not yet launched officially but is very good.

I'd go with a dedicated server unless you really need the expandability that a cloud environment provides - which most people don't.

For those that don't require this expandability, you will find that you are paying significantly more than you would probably be if you simply went with a dedicated server (or even a VPS).

Clouds are incredible and work wonders for some, but for others.. it's just not so suitable.

The idea with Amazon's cloud is that you build an application using multiple smaller cloud instances. Using one big server instance isn't really "cloud". When you have an application that can take advantage of multiple server instances (e.g. load balancing web requests, clustering your database, etc) you can then scale up and down your system based on current load. For instance, you may know that on weekends, you need twice as much capacity as on weekdays. With a well engineered cloud solution, you can automatically provision additional web serving instances and database nodes to services the greater number of requests at the weekend, shutting those nodes down once you're back to Monday.

This means that instead of paying for capacity all the time that you're not using, you only need to pay for capacity when you require it.

As said, however, this requires a better paradigm in application/system design for you to take advantage. If you just need one, big powerful server, then a dedicated server is going to be better for you.

http://www.webhostingtalk.com/showthread.php?t=308055

Here you can find that the performance of ec2 is disappointed at best (in comparison with VPS!).
A colleague mine did a performance test and while the disk is pretty fast but the rest is regular to bad.

One of the advantage of ec2 (apart of scalability) is the amount of free "pre-custom" installation.

Did you account for Amazon's 'Reserved Instances'? If you know you are going to have a full server up for a year, you get a 30% discount (I think). If it is 3 yrs, you get a 50% discount.

Multiple IPs can be handled using their Load Balancer.

But mostly, I agree with dazmanultra. Do you really need one big server? Or are you cramming a bunch of stuff that would naturally run on their own servers onto one big iron?

from my experience, dedicated servers are still the way to go. Amazon is ridiculously bad at disk I/O, we had one c++ script take 40 seconds long to build an index of a database on an EC2 instance, whereas with rackspace cloud servers it takes about 20 seconds, on a dedicated server with a 10k rpm drive, about 5 seconds.

The other issue is the uplink ports. With cloud computing through Amazon or Rackspace, you'll rarely get more than 40 or 50mb uplink. Dedicated servers, either a 10mb or 100mb guaranteed which means you aren't sharing the port speed with anyone.

The thing with cloud computing, while cpu/ram is dedicated, disk IO isn't and the port uplink is NOT (to my knowledge). For us this is a real big problem which is why we are still using dedicated servers for most of our stuff.

It all really depends on your own setup. With videos and bandwidth intensive stuff like that, I'd highly recommend starting with dedicated servers. with cloud computing bandwidth is bloody expensive. You'd be paying around $2,000 for 10TB bandwidth alone when it's included with the $300 fee with singlehop.

But with dedicated servers, hardwares always a problem whereas with cloud computing it isn't.

Yes, Amazon IO is terrible. Some have had success with RAID striping EBS volumes to get respectable bandwidth, but not everyone sees a benefit (apparently sequential writes don't get much better when you RAID over EBS).

But, to really take advantage of Amazon's infrastructure, you might consider more than just EC2. Serving videos out of S3/CloudFront can give you massive scalability benefits...

Some "consultant" then recommended my client towards Amazon EC2 saying its cheaper and better than a dedicated server. The Amazon price calculator (using average transfer rates) for a comparable system gives me a total of $626.92 per month if I use EC2 storage (risky...) and $634.42 if I use S3 storage.

However, if I prototype an entire year (using historical data), the Singlehop server is $3,768.00 per year (just 314x12) - but the Amazon server is a whopping $8,324.16, more than twice as much.

So am I missing something here or are we just below the economy of scale necessary to actually see the cost savings?

You are not missing ANYTHING.

The current generation of "cloud" solutions are nothing more than VPSs that can scale to 1 server exactly. They are frequently plagued with poor latency, i/o issues, bad and inconsistent performance, lots of limitations/restrictions, and high price. The ones that do scale to multiple servers require you to fit your site/software into their framework and platform (and still have all the problems as listed previously).

I would stick to dedicated. It wins hands down.

EC2 was never meant to be cost effective. Its selling point is horizontal scalability. Those who use EC2 need to scale out at any cost. If one dedicated server is enough for the task at hand, then EC2 doesn't add any value to your client. If on the other hand you need to scale out to 5 or more dedicated servers, you can start thinking about the cloud. Although one could easily build a private cloud if needed.

Regards
Joe / UNIXy

Amazon Web Services is not for everyone. It definitely is not the fastest and ideal situation for all businesses. It definitely is not the cheapest.

However Amazon EC2 and Amazon S3 can put some great tools in the hands of business to really help them deploy their infrastructure as it should be and provide a quality of service.

In some situations it can be cheaper, but others you can realize you were spending too little on your infrastructure.

It offers load balancing, version control, instance scaling and other services that would just be too expensive to deploy in-house.

Business should get to know and understand Amazon Web Services just beyond the cloud hype. See what truly works for their business and see where there is a good fit.

I've taken on a client with a news website that serves both static and video (http, not streamed) content, plus a couple of newsletters that are sent out regularly. After a few rounds of briefs and comments, I had specced out a server on Singlehop.com with the following characteristics:

.- Core2Quad @ 2.83Ghz
.- 6.00GB RAM
.- 1TB Hard disk
.- 5 Fixed IPs
.- 10TB Monthly transfer

, at a total cost of $314.00 monthly.

Some "consultant" then recommended my client towards Amazon EC2 saying its cheaper and better than a dedicated server. The Amazon price calculator (using average transfer rates) for a comparable system gives me a total of $626.92 per month if I use EC2 storage (risky...) and $634.42 if I use S3 storage.

However, if I prototype an entire year (using historical data), the Singlehop server is $3,768.00 per year (just 314x12) - but the Amazon server is a whopping $8,324.16, more than twice as much.

So am I missing something here or are we just below the economy of scale necessary to actually see the cost savings?

The Benefits. Your dedicated server will more than likely have more uptime, and less problems than with Amazon's EC2. Sorry. There service still has a long way to mature.

I've taken on a client with a news website that serves both static and video (http, not streamed) content, plus a couple of newsletters that are sent out regularly. After a few rounds of briefs and comments, I had specced out a server on Singlehop.com with the following characteristics:

.- Core2Quad @ 2.83Ghz
.- 6.00GB RAM
.- 1TB Hard disk
.- 5 Fixed IPs
.- 10TB Monthly transfer

, at a total cost of $314.00 monthly.

Some "consultant" then recommended my client towards Amazon EC2 saying its cheaper and better than a dedicated server. The Amazon price calculator (using average transfer rates) for a comparable system gives me a total of $626.92 per month if I use EC2 storage (risky...) and $634.42 if I use S3 storage.

However, if I prototype an entire year (using historical data), the Singlehop server is $3,768.00 per year (just 314x12) - but the Amazon server is a whopping $8,324.16, more than twice as much.

So am I missing something here or are we just below the economy of scale necessary to actually see the cost savings?


EC2 will severly disappoint you based on what I have read in performance compared to a dedicated - EC2 has issues with performance from time to time during certain peak times.

On another note - that seems pretty pricey for that dedicated that you are looking at it unless its fully managed.

@sailor I have to say from running 10-20 servers permanently in the cloud for the last 2 years. And I run conference web sites for SAP, Google, and CA there are not the major performance and IO latency issues that people who read and speculate claim.

Sure you are on a virtualized environment. With higher computer instances available to me, and regional data centers to deploy in...it kicks but over my historical rackspace and other colo relationships I've been in.

I had my EC2 budget doubled this year over last year because of performance gains delivered on a CRM system for SAP internal.

I always say Amazon Cloud is not for everyone, but people really need to try out and see what services fit for their business.

Cloud computing is high availability..

Your uptime on the Amazon network will be higher.

@sailor I have to say from running 10-20 servers permanently in the cloud for the last 2 years. And I run conference web sites for SAP, Google, and CA there are not the major performance and IO latency issues that people who read and speculate claim.




sure there are - they are well documented. Just read all about it. You surely cant expect us to believe you as one voice over what is well documented issues all over the web. EC2 is a for testing and development and small instances but not in my opinion for mission critical stable performance.

Your uptime on the Amazon network will be higher.


not than a private cloud built on a HA platform.

Most of the customers I've personally spoke to that came from Amazon complained of there uptime being at about 70%. I think you will get higher uptime with a dedicated server.

Up time on a cloud hosting environment will definitely be higher.

70% up time. This is hyperbole!

As a cloud hosting environment, including EC2, is on a redundant hardware platform, the downtime will be much lesser than a single dedicated server.

Cheers! :)

Up time on a cloud hosting environment will definitely be higher.

70% up time. This is hyperbole!

As a cloud hosting environment, including EC2, is on a redundant hardware platform, the downtime will be much lesser than a single dedicated server.

Cheers! :)

Its really not that simple. I see you are a reseller of EC2 so I understand your fervent support of it. There are many factors that come into play here including the customer base and nature of use of the particular cloud you are on that will affect uptime and usability / performance.

It will also depend on the storage schema they have as well - so being a "go team!" ra ra cheerleader on a subject without getting into details is a bit shallow.

Up time on a cloud hosting environment will definitely be higher.

70% up time. This is hyperbole!

As a cloud hosting environment, including EC2, is on a redundant hardware platform, the downtime will be much lesser than a single dedicated server.

Cheers! :)

While I agree in general up time on a cloud hosting environment SHOULD be higher, it's definitely not always the case. Besides hardware, cloud environments also operate on more sophisticated software layer which as we've seen with the earlier versions of cloud computing platforms that it has had it's fair share of problems. For example, the underlying engine or distribution server not being properly redundant and truly high available causing complete cloud outages of many servers, there's networking concerns, privacy concerns, and still an issue that is plagued among many cloud technologies is disk I/O.

A properly built cloud infrastructure IS better than an individually dedicated server, however so much of that depends on the technology that runs that cloud, the administration, and what is on it. For this reason, Virtual Private Data Center (VPDC) has become popular as it allows the customer to create mini private cloud's, so they retain complete control over their environment and can build it according to their needs so those problems are not a concern.

Solutions like Amazon's may work fine for some, but performance and reliability is something that will continue to plague it due to the nature of the business. It is just not built or managed to the specifications that some other providers do, but that is not Amazon's target audience either. They know what they are doing, and they are no doubt successful at the market they target, but Amazon is surely not a one-size-fits-all for cloud hosting.

Up time on a cloud hosting environment will definitely be higher.


I would say the main cause of downtime is sofware updates and failures; I don't see how a cloud hosting environment can do anything to reduce those outages.

I would say the main cause of downtime is sofware updates and failures; I don't see how a cloud hosting environment can do anything to reduce those outages.

A properly designed cloud application can separate a traditional "dedicated server" into many appliance's, or "vm's" if you may. Redundancy and reliability can in fact go up just with this alone.

Ie: you need to upgrade a web server appliance, you can do one at a time as they can be load balanced through cloud rebooting only one at a time causing no downtime. Or a failure with a single appliance that is redundant also causes no downtime.

Hardware however is of course still a concern. Honestly, in my experience, hardware is still a greater cause for downtime, latency, or general problems than software. A carefully and properly administrated software environment should be very stable.

a Single server solution has a single point of failure. I wouldn't consider it a "cloud" solution. Good for development yes, critical uptime, no.

I think your talking about like a lamp cluster

http://img52.imageshack.us/img52/109/lampm.jpg

A properly designed cloud application can separate a traditional "dedicated server" into many appliance's, or "vm's" if you may. Redundancy and reliability can in fact go up just with this alone.

Ie: you need to upgrade a web server appliance, you can do one at a time as they can be load balanced through cloud rebooting only one at a time causing no downtime. Or a failure with a single appliance that is redundant also causes no downtime.

Hardware however is of course still a concern. Honestly, in my experience, hardware is still a greater cause for downtime, latency, or general problems than software. A carefully and properly administrated software environment should be very stable.

Exactly.

Amazon EC2 provides developers the tools to build failure resilient applications and isolate themselves from common failure scenarios. from http://aws.amazon.com/ec2/

That does not mean Amazon will automatically handle hardware failures by migrating virtual servers to available hosts. It is up to the developer to have multiple instances of their application hosted on various cloud servers and load balanced like cloudweb mentioned. A single virtual server on a cloud host is just as good as a cheap VPS.

Exactly.

from http://aws.amazon.com/ec2/

That does not mean Amazon will automatically handle hardware failures by migrating virtual servers to available hosts. It is up to the developer to have multiple instances of their application hosted on various cloud servers and load balanced like cloudweb mentioned. A single virtual server on a cloud host is just as good as a cheap VPS.

Well I wouldn't say that's entirely true but the term "cloud" is so diluted as a buzz word I don't blame you at all for thinking that way. Amazon for some reason still hasn't figured that one out. While you can have multiple instance's to create TRUE HA (as mentioned above to prevent software concerns), in a properly designed cloud environment even a single VPS can and should automatically restart and run on another physical server in that cloud. It is inherently better than a traditional VPS for several reasons, that being one of them.

Here's a good read on the Disk I/O comparisons of several clouds vs dedicated servers:
http://blog.cloudharmony.com/2010/06/disk-io-benchmarking-in-cloud.html

The guy went really in depth with it.

Its really not that simple. I see you are a reseller of EC2 so I understand your fervent support of it. There are many factors that come into play here including the customer base and nature of use of the particular cloud you are on that will affect uptime and usability / performance.

It will also depend on the storage schema they have as well - so being a "go team!" ra ra cheerleader on a subject without getting into details is a bit shallow.
Wow! It is the first time I have heard/read any one say that cloud hosting on any platform will give an uptime of 70%! That seems like a deep and thoughtful comment!

Cloud hosting platform is built on redundancy - be it any provider. As compared to a single dedicated server, as I mentioned, it will be much better, all other factors withstanding. Clustered architecture provides for the hardware fail over, which will need a number of servers if a customer/user were to try to replicate.

Please note. We are not a reseller of any provider, including the one you mentioned. We only provide third party support on various platforms.

However, kindly refrain from making personal comments on this forum and stick to facts and reasons. Like I have given you reasons why a cloud hosting platform will be more reliable rather than comment that you are saying this because you are not offering cloud hosting.

While I agree in general up time on a cloud hosting environment SHOULD be higher, it's definitely not always the case. Besides hardware, cloud environments also operate on more sophisticated software layer which as we've seen with the earlier versions of cloud computing platforms that it has had it's fair share of problems. For example, the underlying engine or distribution server not being properly redundant and truly high available causing complete cloud outages of many servers, there's networking concerns, privacy concerns, and still an issue that is plagued among many cloud technologies is disk I/O.

A properly built cloud infrastructure IS better than an individually dedicated server, however so much of that depends on the technology that runs that cloud, the administration, and what is on it. For this reason, Virtual Private Data Center (VPDC) has become popular as it allows the customer to create mini private cloud's, so they retain complete control over their environment and can build it according to their needs so those problems are not a concern.

Solutions like Amazon's may work fine for some, but performance and reliability is something that will continue to plague it due to the nature of the business. It is just not built or managed to the specifications that some other providers do, but that is not Amazon's target audience either. They know what they are doing, and they are no doubt successful at the market they target, but Amazon is surely not a one-size-fits-all for cloud hosting.
Amazon or any cloud hosting provider can never be suitable to ALL users. Never stated that. However EC2 or Rackspace or GoGrid or Storm (or any other cloud providers with similar features) do offer distinct advantages over traditional hosting.

For simulating a similar environment one would need to work on a larger scale. However, a cloud provider these advantages at price ranges of a dedicated server.

Cloud hosting platform is built on redundancy - be it any provider. As compared to a single dedicated server, as I mentioned, it will be much better, all other factors withstanding. Clustered architecture provides for the hardware fail over, which will need a number of servers if a customer/user were to try to replicate.


I don't understand cloud computing.

Take a simple example; suppose I have a web application, which runs on a dedicated server.

A user signs in and establishes a session; that means the application can verify that subsequent requests from that user are indeed from that user; it can maintain state information somewhere within the server; and it can use that state information to control what the application delivers to the user in subsequent requests. I think this is reasonably typical of a proportion of web applications today.

How does that work in a cloud environment? Do subsequent requests from the same user go to the same - what? Does the application have to be redeveloped to cope with maintaining user sessions in a cloud environment?

I don't understand cloud computing.

Take a simple example; suppose I have a web application, which runs on a dedicated server.

A user signs in and establishes a session; that means the application can verify that subsequent requests from that user are indeed from that user; it can maintain state information somewhere within the server; and it can use that state information to control what the application delivers to the user in subsequent requests. I think this is reasonably typical of a proportion of web applications today.

How does that work in a cloud environment? Do subsequent requests from the same user go to the same - what? Does the application have to be redeveloped to cope with maintaining user sessions in a cloud environment?
1. From the application developer's stand point, little changes when you move from a traditional hosting environment to a cloud hosting environment.

2. In simple words, you will get a cloud server (or instance) with some specifications (say, 2 GB RAM / 100 GB disk space ). To this will be attached one or more IPs. Then you go ahead and install applications and run them in pretty much the same way that you do on a physical dedicated server.

3. The server or instance is recognized with IPs, just as they are for a dedicated server. As a developer you get root access to your OS - be it Linux or Windows and all other privileges. You can install whatever software you need, over this - Webmin / Plesk / Vnc...

4. If you use one cloud instance then all subsequent user requests will go back to the same IP and therefore the server/instance.

If you use multiple instances by way of a load balancer then you can define session persistence, if needed, in which case the user will go back to the same instance, while still in a session.

Hope it answers your Q.

Cheers!

1. From the application developer's stand point, little changes when you move from a traditional hosting environment to a cloud hosting environment.

2. In simple words, you will get a cloud server (or instance) with some specifications (say, 2 GB RAM / 100 GB disk space ). To this will be attached one or more IPs. Then you go ahead and install applications and run them in pretty much the same way that you do on a physical dedicated server.

3. The server or instance is recognized with IPs, just as they are for a dedicated server. As a developer you get root access to your OS - be it Linux or Windows and all other privileges. You can install whatever software you need, over this - Webmin / Plesk / Vnc...

4. If you use one cloud instance then all subsequent user requests will go back to the same IP and therefore the server/instance.

If you use multiple instances by way of a load balancer then you can define session persistence, if needed, in which case the user will go back to the same instance, while still in a session.

Hope it answers your Q.

Cheers!

Thanks; I understood that! So it's one or more virtual servers.

How do I plan the I/O performance?

On a dedicated server, I know the disk configuration and disk spec, I know the program I am designing, so I can ensure that the combination of program design and disk subsystem will meet my performance targets.

For example, I can design a transaction that requires 30 random disk IOs, and know I can meet a target transaction time of 1 second using a pair of 7.2K RPM disks with peak-hour load of 2 transactions per second.

When using a cloud, how can I ensure my application design meets my performance target?

Thanks; I understood that! So it's one or more virtual servers.

How do I plan the I/O performance?

On a dedicated server, I know the disk configuration and disk spec, I know the program I am designing, so I can ensure that the combination of program design and disk subsystem will meet my performance targets.

For example, I can design a transaction that requires 30 random disk IOs, and know I can meet a target transaction time of 1 second using a pair of 7.2K RPM disks with peak-hour load of 2 transactions per second.

When using a cloud, how can I ensure my application design meets my performance target?

With any VPS or Cloud solution, you're not going to have guaranteed IOPS. All virtual servers will be sharing the same set of disks and competing for IO. This is one of the advantages of a dedicated server. There are bare-metal clouds that take dedicated servers and throw an API and usage based billing in front of it.

In the cloud it would be best to cache as much as you can using memcached, ehcache, etc.. and distribute DB writes across replicated servers.

With any VPS or Cloud solution, you're not going to have guaranteed IOPS. All virtual servers will be sharing the same set of disks and competing for IO. This is one of the advantages of a dedicated server. There are bare-metal clouds that take dedicated servers and throw an API and usage based billing in front of it.

In the cloud it would be best to cache as much as you can using memcached, ehcache, etc.. and distribute DB writes across replicated servers.

...unless the customer is utilizing an entire server on a cloud. If he is in fact needing an entire dedicated server, then one could assume he could justify the use of an entire servers worth of resources on a cloud environment, therefore not necessarily competing for IO in a local cloud storage type of solution.

EC2 has a lot of nice features like autoscaling, load balancing, automatic duplication of EBS volumes, and a better pricing (hourly) and deployment (automatic provisioning) model for bursty services. However, they do limit to single IP address and don't provide RDNS capabilities. Their user interface is also catered to more technical users. They have 3 pricing options, including standard hourly, discounted reserve pricing, and also spot pricing. The latter two can significantly reduce your cost. Spot pricing for example, is often 50% or more less than regular hourly. If you use the full hourly price for you max bid on a spot instance, you can generally run that instance indefinitely while paying significantly less.

You are funny. I dont think I need you to give me lessons on how to post on this forum. BTW - welcome to the forum - I see you have been here for 3 months.

I do offer cloud hosting in case you were wondering to clear the air on that - I am agnostic and only speak of facts. The fact is that the uses of the cloud and the customer base DO highly affect the stability and performance of the cloud. That is why enterprise business is mainly focusing on private cloud offerings for delivery and public cloud in the same provider for test and development in a cheaper format.



Wow! It is the first time I have heard/read any one say that cloud hosting on any platform will give an uptime of 70%! That seems like a deep and thoughtful comment!

Cloud hosting platform is built on redundancy - be it any provider. As compared to a single dedicated server, as I mentioned, it will be much better, all other factors withstanding. Clustered architecture provides for the hardware fail over, which will need a number of servers if a customer/user were to try to replicate.

Please note. We are not a reseller of any provider, including the one you mentioned. We only provide third party support on various platforms.

However, kindly refrain from making personal comments on this forum and stick to facts and reasons. Like I have given you reasons why a cloud hosting platform will be more reliable rather than comment that you are saying this because you are not offering cloud hosting.

With more awareness, education & benefit realization, cloud hosting is finding rapid and widespread adoption, across sectors and business sizes.

With more awareness, education & benefit realization, cloud hosting is finding rapid and widespread adoption, across sectors and business sizes.

agreed - I think it is under adopted.

agreed - I think it is under adopted.

+2. :agree: