The Custom-Made Cloud

November 10, 2010

Not to toot my own horn, but I am an actual Rocket Scientist (well an Aerospace Engineer, but Rocket Scientist sounds way cooler). When you are a Rocket Scientist, most of your time is spent in dealing with facts – universal constants, formulas, and a data set that has been validated countless times over. My role at the CTO at SoftLayer is sometimes a challenge because I have to deal with the terrific hyperbole that the tech world inevitably creates. Consider the Segway, Unified Messaging, etc. I think that cloud computing has also fallen prey.

The cloud promises a lot and it does deliver a lot.

  • Control puts decisions and actions in the hands of the customer. Self-service interfaces enable automated infrastructure provisioning, monitoring, and management. APIs provide even greater automation by supporting integration with other tools and processes, and enabling applications to self-manage.
  • Flexibility provides a broader range of capabilities and choices, enabling the customer to strike ideal balance of capital and operating expenses. In addition, access to additional infrastructure resources happens in minutes rather than week enabling you to respond "on demand" to changes in demand.
  • Flexibility and control combined give administrators more choice. Who manages infrastructure (Internal staff or service provider?) Where are workloads processed in an internal datacenter or in the public cloud? When are workloads processed – is this resource-driven or priority-driven? How much to consume – is this policy-driven or demand-driven? How is IT consumed – via central administration or self-service?

Despite its numerous benefits, the operational and cost effectiveness for many enterprises is challenged by the fact that most cloud services come in limited configurations and only serve as standalone solutions. One cloud does not fit all – Fixed specs do not allow administrators to optimize a cloud environment with the ratio of processing power, memory or storage that its intended application needs for its best performance. Most cloud service providers offer a relatively small number of preconfigured virtual machine images (VMI), often starting with small, medium and large VMIs, each with preset amounts of CPU, RAM and storage. The challenge is that even a few sizes (versus only one) don't fit everybody's needs. For example, applications perform best when they are running on servers with optimized configurations. And every application has unique resource demands. If the server is "too small," performance issues may arise. If the server is "too large," the customer ends up paying for more resources than necessary.

To a degree we have already been doing lots of "cloudy" things given our focus on automation. Combine that with a set of tools that let customers self-provision and I think you see where I am headed. The next step up the value chain is SoftLayer's "Build Your Own Cloud" solution. It delivers all of the benefits that I discussed above, but adds the logical step of handing configuration control to the customer. Customers are able to determine a number of things about the environment that their cloud sits on.

Cloud Computing Options
Cloud Computing Options Part Two (Monthly)

The end result is a cloud environment that is fit for customer purpose and customer cost. A classic win-win situation. I wonder what we will think of next.

-@nday91

Comments

November 25th, 2010 at 4:24pm

I would agree that SL's offers above-average flexibility for cloud instances in terms of virtual hardware configuration. However, SL still loses badly to Amazon EC2 when it comes to OS template flexibility, which can often be more important to a customer than virtual hardware choices. You could fix this shortcoming in a hurry if you would just implement RescueLayer for CCI. For example, if I wanted to set up a dedicated server with the Linux system on an LVM partition, I can do it today without making a special request and waiting for a support person. I've got no complaints about SL's support crew, but they're not instantaneous like automated provisioning. I would simply order the server with standard partitioning, then boot in RescueLayer to repartition it myself. With SL's current cloud offering, I can't do the same. Can I maybe build a paravirtualized VM image on my own local Xen and upload it to my cloud template storage? Nope. Is there perhaps a way to temporarily connect one CCI's primary disk as a secondary disk on another CCI (while the first CCI is shutdown)? Nope. Best I can do without a ticket is to settle for LVM on a second virtual disk and organizing my filesystem mount points to compensate. Amazon EC2 allows me to do a zillion things using their AMI templates without opening tickets and waiting for somebody. I don't realistically expect SL to implement the same level of template flexibility, but if you would implement RescueLayer, that would let me handle everything on my own.

Leave a Reply

Filtered HTML

  • Web page addresses and e-mail addresses turn into links automatically.
  • You can enable syntax highlighting of source code with the following tags: <pre>, <blockcode>, <bash>, <c>, <cpp>, <drupal5>, <drupal6>, <java>, <javascript>, <php>, <python>, <ruby>. The supported tag styles are: <foo>, [foo].
  • Allowed HTML tags: <a> <em> <strong> <cite> <blockquote> <code> <ul> <ol> <li> <dl> <dt> <dd>
  • Lines and paragraphs break automatically.

Plain text

  • No HTML tags allowed.
  • Web page addresses and e-mail addresses turn into links automatically.
  • Lines and paragraphs break automatically.
By submitting this form, you accept the Mollom privacy policy.

Comments

November 25th, 2010 at 4:24pm

I would agree that SL's offers above-average flexibility for cloud instances in terms of virtual hardware configuration. However, SL still loses badly to Amazon EC2 when it comes to OS template flexibility, which can often be more important to a customer than virtual hardware choices. You could fix this shortcoming in a hurry if you would just implement RescueLayer for CCI. For example, if I wanted to set up a dedicated server with the Linux system on an LVM partition, I can do it today without making a special request and waiting for a support person. I've got no complaints about SL's support crew, but they're not instantaneous like automated provisioning. I would simply order the server with standard partitioning, then boot in RescueLayer to repartition it myself. With SL's current cloud offering, I can't do the same. Can I maybe build a paravirtualized VM image on my own local Xen and upload it to my cloud template storage? Nope. Is there perhaps a way to temporarily connect one CCI's primary disk as a secondary disk on another CCI (while the first CCI is shutdown)? Nope. Best I can do without a ticket is to settle for LVM on a second virtual disk and organizing my filesystem mount points to compensate. Amazon EC2 allows me to do a zillion things using their AMI templates without opening tickets and waiting for somebody. I don't realistically expect SL to implement the same level of template flexibility, but if you would implement RescueLayer, that would let me handle everything on my own.

Leave a Reply

Filtered HTML

  • Web page addresses and e-mail addresses turn into links automatically.
  • You can enable syntax highlighting of source code with the following tags: <pre>, <blockcode>, <bash>, <c>, <cpp>, <drupal5>, <drupal6>, <java>, <javascript>, <php>, <python>, <ruby>. The supported tag styles are: <foo>, [foo].
  • Allowed HTML tags: <a> <em> <strong> <cite> <blockquote> <code> <ul> <ol> <li> <dl> <dt> <dd>
  • Lines and paragraphs break automatically.

Plain text

  • No HTML tags allowed.
  • Web page addresses and e-mail addresses turn into links automatically.
  • Lines and paragraphs break automatically.
By submitting this form, you accept the Mollom privacy policy.