Microservers are one of the hot topics today. As always microserver vendors are rolling out their products to the market and the businesses are trying to see what advantages these servers will bring to their data centers. When considering an investment for a server that has low power, low consumption and low space requirements, businesses do want to know exactly what they are getting for their bucks.
First and foremost, microservers do not have a clear definition of what they are. As Kevin Huiskes of Intel puts them “any server with a large number of nodes, usually with a single socket or multiple low-power processors and shared infrastructure.” Broadly we can define them as low-power (primarily in terms of computing power), high-density servers which are designed for specific, scale-out workloads.
Scale-out workloads are tasks that can be executed in parallel, such as serving static HTML pages. Following this example, serving a static HTML page does not require too much of a computing power. But when you need to serve hundreds of thousands of item displays in a second (consider all the elements in a page, backgrounds, images, text etc.) the computing power required to undertake is significant. But you do not need one expensive, high-power server to do this either. You can split the serving task to multiple low-power servers and scale out your task. As the task at hand grows, you can add further servers.
When this is the case, of course blade centers are the first servers to take the hit. Comparing blade servers to microservers, blade servers are high power, high consumption servers requiring more cooling. If the infrastructure requirements do not justify such a computing power, such as delivering virtual desktop infrastructures, then why not consider microservers in place of blades?
There is also another advantage that microservers bring, which is the pooling their resources. With shared processing and clustering, microservers can achieve more demanding workloads. This capability makes them an ideal candidate for data analytics where the computing requirements can be distributed or consolidated as necessary.
This consolidation also brings two more scenarios. The first scenario is using the microservers for the cloud infrastructures. Imagine an IaaS scenario where you need to spin off a couple of servers where some testing scenarios will run. Imagine another scenario where you provide an SaaS offering – say an email service – where significantly low computing power is required per user. These are just two scenarios that I can think of now in a flash where microservers would be an excellent fit.
The next question is the media storage. Media storage and serving on the network, whether an Inter-, Intra- or Extranet setting, similarly do not require significant horsepower per user, making microservers an excellent candidate for the task.
Another point to keep in mind about the microservers are the storage requirements. Microservers offer both internal storage and shared storage on the shared-infrastructure chassis. Your storage requirements is a point to consider with the microservers: whether the storage will be sufficient for the space and/or the I/O.
Finally, there is a balance between the traditional servers and the microservers.Although Your PC Universe says that microservers use %85 less power and require %63 less cost to setup, there is no point in rushing out and replacing all the servers with microservers. Cost savings look appealing on the paper but businesses need to see the level of performance they need and what amount of this performance can be loaded on the microservers. Without establishing this balance, there will be savings but at the cost performance.
- Featured image: http://www.dell.com/us/business/p/poweredge-c5125/pd