Is Storage + Dense Server Considered Hyperconverged?
What is and what isn’t considered a ‘hyperconverged’ infrastructure?
Question: Is a [Storage Vendor Name] + dense server implementation close enough to be considered hyperconverged?
During the ActualTech Media Introduction to Hyperconvergence webinar held on September 18, 2014, an audience member asked whether the combination of storage from a specific vendor plus a dense server pool could be considered hyperconverged.
On this specific question, the answer is No. Although there will likely be a number of vendors from different specific resource areas – storage, compute, etc. – that attempt to attach their brand to the hyperconvergence moniker, there are some characteristics that must be present in order to a solution to be considered hyperconverged.
Single (Or Massively Simplified) Administration
Hyperconverged infrastructure vendors tout simplified administration of the data center as one of the primary benefits of the solution. Many storage vendors today have integrated big chunks of storage management needs right into existing tools such as vCenter. In that way, those storage solutions are actively working to simplify the data center administration paradigm, which is a great benefit to these solutions.
However, even while many do integrate the various tools into one console, there is still a need to manage underlying storage constructs, such as LUNs. Many hyperconverged systems eschew the need for even these kinds of efforts. As such, there is still a need to manage separate resources in at least some fashion. Again, this may not be true for all storage vendors, so check with your specific vendor before assuming this is true for yours.
Linear Resource Scalability
Even if the software is completely and 100% simplified with a specific storage vendor, there remains the need to separately manage both storage and compute. Hyperconvergence makes the basic assumption that these two resources re tightly coupled with the storage resource being a series of locally attached disks which the vendor’s software layer then aggregates and presents to the cluster as single unified data store.
Any solution that uses standalone storage is not hyperconverged. With a hyperconverged solution, every time a new node is added, there is additional compute, storage, and network capability added to the cluster. Simply adding a shelf of disks to an existing storage system does not provide linear scalability and can eventually lead to resource constraints if not managed carefully.
For more information, please download my book entitled Hyperconverged Infrastructure for Dummies in which I discuss why linear resource scalability is one of the key benefits of hyperconverged infrastructure.