Hyperscale data centers have architectures that are designed to provide a single, massively scalable compute architecture. The architecture is typically made up of small, individual servers, called nodes, that provide compute, storage and networking. These nodes are then clustered together and managed as if they were a single entity. Nodes are typically deployed from inexpensive, off the shelf servers.
Hyperscale designs have been made popular by Web 2.0 companies that need the ability to rapidly morph from industry start up to industry juggernaut. They’ve also become popular with large providers that play host to many of these same startups. Now, however, the Hyperscale Data Center is beginning to attract the attention of the more traditional enterprise environments which are experiencing many of the same “startup to scale” challenges that Web 2.0 companies have to contend with.
Start Small But Scale Wide
The idea behind building a hyperscale architectures is to start small in order to keep upfront investments as low as possible. Then as demand grows, the infrastructure should be able to expand simply by adding nodes to the cluster. Most, if not all the scale out software used in this market is specifically designed make node aggregation as seamless as possible. This is an ideal scaling model for subscriber driven organizations because they can grow the data center at the same pace as they add customers. For more traditional enterprises, while not subscriber driven, the appeal is similar, as an application or initiative grows, just add more nodes. Both scenarios derive the dual benefit of a cost effective initial implementation without the scaling limitations of a more traditional entry level solution.
Automated Load Distribution
When nodes are added, the hyperscale software, in most cases, will then re-position or auto-balance workloads to the newly added nodes. Some hyperscale clustering software will allow for certain workloads to obtain resource priority. That way performance guarantees can be established for applications that are bound by stringent service level agreements (SLAs). In this type of hyperscale software, the maximum level of compute and memory available is limited to the specific node that the workload is running on. In this type of configuration, compute and memory are not shared across the cluster.
On the other hand, there are other hyperscale software solutions that actually share compute and memory; similar to how a clustered application is configured. This leads to a more tightly integrated cluster where the available compute and memory resources of all the nodes can be combined together.
Hyperscale Storage
Generally, storage has been a major challenge for these architectures. Ideally, storage is installed in each node as it is added to the cluster and is then pooled across all the nodes by the hypervisor storage software. Essentially as more nodes are added, performance and capacity increases linearly. If more capacity or more performance is needed simply add more nodes. The problem with this approach, however, is that each time the cluster is expanded to increase storage capacity, valuable compute resources are needlessly wasted.
As a result, hyperscale architectures are moving to solid state disk (SSD) to solve these problems. The challenge with this approach is most SSDs don’t have the storage density to meet the capacity demands of most large scale environments. Now, expensive SSDs are populated in the nodes, to meet a capacity demand. Due to these issues, many hyperscale architectures have to resort to “split-clusters” – isolating performance clusters from capacity clusters. The problem is that this approach reduces the operational effectiveness of the environment and is still not a cost efficient way to scale the environment.
Another alternative is to use PCIe SSDs. This is something that many of the larger sized hyperscale data centers have done. These devices have the density and the performance to meet both demands but at an added cost. There is also an added layer of complexity since many of the PCIe SSD offerings do not leverage existing storage protocols for access nor can the node boot off them. The larger, more established hyperscale data centers often wrote the software code that drives their hyperscale environment and as a result, they have the ability to make the customizations necessary to support and take full advantage of native PCIe SSDs.
While PCIe SSD often makes sense for these environments, an alternative is needed that can provide both capacity and performance. Additionally, for those organizations that don’t possess the software code, the storage medium should be accessible through traditional storage protocols like those used by standard hard disk drives.
The good news is that, thanks to the latest generation of SAS connectivity and new levels of storage density, drive form factor SSDs are able to address this demand. For example, SMART Storage recently announced a SAS based 2TB 2.5″ SSD. This type of solution is ideal for hyperscale data centers that can not afford the cost of PCIe SSD or don’t have the ability to customize their applications to take advantage of native PCIe SSD performance.
SMART Storage is a client of Storage Switzerland
George Crump is the Chief Marketing Officer at VergeIO, the leader in Ultraconverged Infrastructure. Prior to VergeIO he was Chief Product Strategist at StorONE. Before assuming roles with innovative technology vendors, George spent almost 14 years as the founder and lead analyst at Storage Switzerland. In his spare time, he continues to write blogs on Storage Switzerland to educate IT professionals on all aspects of data center storage. He is the primary contributor to Storage Switzerland and is a heavily sought-after public speaker. With over 30 years of experience designing storage solutions for data centers across the US, he has seen the birth of such technologies as RAID, NAS, SAN, Virtualization, Cloud, and Enterprise Flash. Before founding Storage Switzerland, he was CTO at one of the nation's largest storage integrators, where he was in charge of technology testing, integration, and product selection.
StorageSwiss.com - The Home of Storage Switzerland 