Best uses for servers that have been replaced with a hyper-converged environment
If you haven’t already implemented a hyperconverged infrastructure, you are at least thinking about trying it out. After making the switch, you will have an existing infrastructure that probably consisted of a traditional 3-2-1 or “inverted pyramid” architecture that will be replaced with your shiny, new HC infrastructure! Things couldn’t be better, but that leaves you with some old equipment that it would be handy if you could continue to use in some way. This might include servers and almost certainly includes storage either NAS or SAN. You don’t want that to go to waste. And having a good reuse plan makes cost justifying a new hyperconverged infrastructure that much easier.
Storage of all kinds is trivial to re-deploy. Adequate storage for backups and archives is almost always in short supply or difficult to justify. Reusing primary storage for backups normally means higher performance for backups and restores than is normally available. In many cases, it also means more reliable hardware than is often chosen for backups.
Instead of the backup infrastructure being treated as a second citizen within the organization, reuse may buy many years of highly reliable and performant backup and/or archival storage. As an archival system, this can offload capacity from the primary HC infrastructure lowering cost of acquisition, as well.
In cases where backup or archival capacity are not useful, storage almost always has a place in labs or for non-critical workloads. Gaining the ability to have testing storage can be a huge boon to nearly any environment.
Re-purposing compute capability may be even easier. If even more storage is needed or warranted than the reused primary storage can provide, most compute can be reused as additional storage quite easily as well. Or compute nodes might be integrated into some hyperconverged infrastructures, depending on many factors. This, however, would be very rare.
Probably the most common valuable use case for re-purposing compute nodes that have been made redundant through the introduction of a new hyperconverged platform is to build a testing, staging or lab environment. The use of a lab for testing new software, ideas, patches and so forth can be extremely valuable but rarely gets direct funding. These soft benefits are often difficult to sell to a business when they require budgeting but when used to offset the cost of a shiny, new infrastructure can look very attractive.
In larger or more unique organizations it is possible to dedicate retired compute capacity to roles such as decision clusters that would otherwise present a very specific workload need and drain on the primary HC platform. Dedicated Hadoop style business intelligence may be shunted to the older compute nodes, for example.
Core infrastructure on retirement could also be moved to remote or branch offices to provide services at those locations such as local file serving, caches, proxies or active directory services.
It is common to use non-cluster hardware for roles such as the backup server itself (not the storage, but the head unit) and this is yet again a great role for re-purposed compute nodes.
When new HC infrastructure is introduced there is almost always a concern around the sunk cost of the existing equipment which is almost never completely ready to be retired, but there are numerous ways that compute and storage capacity can be put to good use in your shiny new network. There is no need to see the old equipment as “lost”, but rather as additional opportunity to further improve your network, provide additional services and invest in use cases that often go overlooked and underserved.
Of course, all of this ignores the very obvious possibility of simply selling the used equipment and moving on.
Guest Post: Scott Allan Miller