4sysops in the cloud - The reasons

In the last post in this series, I outlined why costs savings were not the reason for moving 4sysops to the cloud. On the contrary, I could have saved money if I had chosen a dedicated server. So what were the reasons? Another argument you hear often is that cloud computing scales better than other environments.

Contents of this article

Cloud scalability ^

In the first post of this series, I described how I figured out that the Amazon's small EC2 instance was not powerful enough to host 4sysops. The good thing about EC2 is that I could easily launch a more powerful virtual machine. If I had ordered a new server at a conventional provider, for example, and realized that the machine was too small, then things would have been a lot more complicated. With EC2, I simply launched a second, more powerful, instance of my private AMI (OS image) and moved the IP address to the new instance. Thus, it didn't cost me extra work time to upgrade to a bigger machine. Furthermore, moving the IP to the new instance was only a matter of seconds, so there was no downtime.

I am now using a High-CPU Medium instance that has five EC2 Compute Units (two virtual cores with 2.5 EC2 Compute Units each). This machine should be about five times as fast as the small instance type I used at first. It also has only 1.7 GB RAM, but that is enough for the 500 page views per hour that my server has to deliver at peak times. So far, the performance is sufficient to run 4sysops easily. However, whenever I start a bundling process (create a new image), the virtual machine starts sweating again.

Hence, cloud computing clearly as an advantage over traditional server hosting when it comes to scalability. The same applies to on-premise computing. If you figure out that the servers you bought last year do not have enough power anymore, you certainly have a problem. And if your servers are too big, that means wasted money.

Unfortunately, if I ever again need a more powerful server, the upgrade won't be that easy. The instance type I am using now is 32-bit, and the next powerful instance type is 64-bit. Hence, I would have to install everything from scratch. Besides, this instance would cost four times as much as the one I have now, which would be much too expensive. Amazon offers only five different machines types.

Technically, cloud computing can offer better scalability than traditional server hosting or on premise computing. It is just that Amazon doesn't really use that advantage. They definitely need more different instance types.

Furthermore, it is not possible to upgrade reserved instances. Many conventional server hosters allow you to upgrade to a bigger server before the end of the term. But not Amazon. This is certainly a major disadvantage of EC2.

Cloud flexibility ^

So why Amazon EC2? The big plus is its flexibility. Let's say, I want to make some major changes on my server. Instead of working on a productive server, I can launch a copy of my instance and mess with it without having to worry that my server will go down if I make a mistake. Once everything is finished, I can terminate the old instance and transfer the IP to the new machine.

Of course, you’ll have similar features if you work with server virtualization in your own datacenter; however, you’ll then need extra computing power for your test machines or short term projects. Usually, when you need a new virtual machine to test something or for a productive server, you will realize that the other admins in your organization have already occupied all the available capacity on your servers. Then, you will start thinking if you could reduce the capacity of some virtual servers or move them to other hosts, or just order new hardware. This costs time.

In the cloud, you simply launch a new test machine and pay only for the time you actively work on it. It simply doesn't matter on which host the instance is running. The location of your virtual machines is virtualized; it doesn't matter anymore. Cloud computing basically is location virtualization.

During off hours you can shut down the instance and launch your configuration whenever you need it again. You don't pay for your instance during this time. The storage costs for the AMI (OS image) are more or less negligible for most machine types. And if you are done with your testing, you purchase a reserved instance and your productive server is online within seconds. You don't have to spend time for ordering the server, wait until it is delivered, mount into the rack, etc.

You also don't have to plan what hardware you’ll need for the next year or so. If you need new capacity, you can launch another instance with a mouse click. And best of all, you can just focus on the software part and don't have to bother with any kind of hardware management.

Thus not cost savings make cloud computing attractive. It just simplifies server provisioning. You can launch new server instances very quickly which gives your company a significant competitive advantage. Therefore, cloud computing does not reduce costs, but it can help your company to increase earnings because the improved flexibility increases the agility of your business.

The flexibility factor was the one I underestimated the most when I heard about cloud computing for the first time. However, I had another reason for moving 4sysops to the cloud. After having more practical experience with EC2, I believe cloud computing has a great future. A lot more competition is coming soon, which will reduce prices, significantly; in fact, there could come a time when cloud computing is cheaper than on-premise computing. As an IT pro, you should be prepared for that time!

Want to write for 4sysops? We are looking for new authors.

Read 4sysops without ads by becoming a member!

1+
Share
3 Comments
  1. John 11 years ago

    Hi Michael, I often read your blog and thought I would give you (and your readers) some feedback on some hands on experience with the EC2 cloud.

    In a word - POOR! Very weak server.

    A client of ours was serving 50,000 uniques running a quad core xeon 4GB ram Dell dedicated server.

    Moved to EC2 and the highest level server Amazon offer and guess what? Exact same config, and with just 1000 uniques it dies. Completely dies under the load of just 1000 visitors. Can you believe that?

    So, if you fall for this EC2 trap, plan clustering. Yes, EC2 is only useful if you plan getting a massive load that a powerful dedicated cannot handle. Use EC2 for clustering 20+ servers, otherwise go dedicated or VPS.

    PS - The S3 service is a different story. Now that is a great services.

    Regards and keep up the blogging.

    1+

  2. Which CMS were you using? My High-CPU Medium Instance can handle 600 visits easily. I suppose it wouldn't have any problems with 1000 either. The High-CPU Extra Large Instance should be 4 times as powerful. Did you use this one?

    1+

  3. John 11 years ago

    Hi Michael,

    Google analytics reports it was 2000 simultaneous visitors that brought the media rich site down running on a high end EC2 instance. With a more even spread of visits over the day it's stable.

    Regards.

    1+

Leave a reply

Your email address will not be published. Required fields are marked *

*

© 4sysops 2006 - 2020

CONTACT US

Please ask IT administration questions in the forums. Any other messages are welcome.

Sending

Log in with your credentials

or    

Forgot your details?

Create Account