February 28, 2014 | Written by: Alexei Karve
Share this post:
When we think of cloud, two words quickly come to mind with respect to resources: virtualized and shared. We also immediately think of a scale out model—horizontal scaling. If you want to handle more load, add more machines. What about scale up? With SoftLayer, that is not an obstacle anymore. There is a choice of physical and virtualized servers, a choice of dedicated or mixed-resource sharing and flex image service for migration. SoftLayer provides granular control during provisioning and the ability to resize currently provisioned servers.
Clouds usually provide a fixed set of sizes called flavors or compute sizes. These start with small, medium or large and come with a fixed number of CPUs, RAM and disk size. With SoftLayer, you are no longer limited to the fixed set of choices. Instead, you get granular control on the choice of these resources. You can select up to 16 CPUs, 48GB of RAM and multiple disks, each with a wide variety of choices on local or storage area network for CloudLayer Computing Instances (CCI). Similarly for bare metal servers (BMI), eight choices are available with up to 16 cores and 64GB of RAM. The widest choice is for dedicated servers with different CPU types, up to 40 cores, 512GB of RAM and a choice of up to 24 SATA or SSD disks as of this writing.
Traditionally, if you made the wrong choice, you may have ended up with a server that was too small and could not handle the load, or a server that was too large and resources went underutilized. With SoftLayer, you can resize your servers and scale up or down with resources based on need. This gives immense flexibility and control to achieve the right balance between virtual servers and physical servers. Enterprise applications that are designed with multiple tiers can have independent requirements and configurations. One tier can scale up or down while the other may scale out or in.
SoftLayer provides another benefit with flex images, allowing you to move between virtual and physical. Customers can therefore start small with cheaper virtual servers and rapidly and easily move to dedicated servers. Thus migration of applications from slow virtual servers to larger dedicated servers allows easy upgrades to faster disks, more memory and CPUs. This service allows customers to copy and store an image of a cloud or dedicated physical server, and then redeploy the image on either type of computing environment.
A virtual server does not usually perform like a physical server. However in most deployments, physical servers often remain underutilized. SoftLayer provides you the ability to use a combination of physical and virtual servers that are on the same VLAN. It allows you to make adjustments by switching and resizing servers for changing workloads. This is like a hybrid electric vehicle that combines a conventional internal combustion engine propulsion system with an electric propulsion system to take advantage of maintenance, distance travelled and speed of refuelling. The flexibility puts the choice in the hands of the customer to find the right balance between physical and virtual servers.
Primary database services are often deployed on physical hardware, which is heavily utilized. Systems consistently running at high utilization, high network load, major storage needs and analytics workloads requiring high I/O are usually provisioned on physical nodes. Adding redundant physical servers costs more. However it may be required for disaster recovery of databases. Web servers that can be horizontally scaled and require high availability are usually provisioned on virtual servers. If a virtual server goes down, you can quickly bring up another. There is usually a base minimum set of required servers that are ordered on a monthly basis. This could be a mix of virtual and physical. For handling higher demand, you can order hourly resources. This can be done daily where you start up servers in the morning and take them down in the evening. For monthly workloads such as payroll processing, you could order servers for only few days of the month. Also for seasonal workloads, you can order servers only for a month. When a testing cycle begins, there will be high demand for servers that are used for a short period of time. You only pay for what you use.
Achieving harmony with optimal cost is thus possible because of the fine-grained control and choices available with a hybrid approach to the cloud infrastructure needs from SoftLayer. What do you think? You can follow me on Twitter @aakarve.