Storage technologies have been constantly evolving over the past few decades, from dedicated application resources, through JBOD arrays, to the latest innovations in space saving and performance such as thin provisioning, caching using SSD or RAM, and data deduplication. This technological development has lent itself to the requirements of large data center environments, yet largely ignored the challenges faced by organizations deploying storage at branch or remote locations.
In this five part series we take a look at how storage solutions have evolved and how data storage vendors are finally addressing the challenges faced by distributed organizations in serving their remote sites with cost-effective and highly available shared storage. Last time we discussed how using dedicated servers per application had deficiencies relating to high availability, scalability, and resource starvation. In part two of this series we look at the subsequent introduction of shared JBOD (Just a Bunch Of Disks) storage to alleviate some of these issues.
The shared JBOD storage array allowed for better disk capacity scaling and increased performance. It made clustering and high availability possible, enabling applications to failover to standby servers in the event of a failure, minimizing application downtime. With JBOD arrays, the available disk capacity was aggregated into storage pools that allowed for better storage capacity utilization.
Some shared storage arrays also provided basic operations such as RAID protection and snapshots.
Limits of the JBOD storage architecture
While clustering and high availability was possible with shared JBOD storage, cluster sizes were limited by the physical connectivity of the JBOD arrays, restricting the number of servers that can be connected to the array. In addition, it is not possible to isolate storage resources between different applications, leading to storage contention and performance bottlenecks.
Due to these issues, “islands” of storage appeared, each island dedicated to workloads with similar performance and protection requirements or to a single department to ensure that performance issues were minimized or localized to a single team. Like the previous architecture, advocating dedicated server and storage resources, these “storage islands” increased management overheads and allowed non-standard technologies to be introduced.
Does this approach address distributed enterprise challenges?
Although this approach has some issues, these simple shared JBOD storage solutions are relatively inexpensive and easy to deploy making them suitable for remote sites with minimal server requirements. However, the inclusion of the shared JBOD array increases the infrastructure footprint (space, power, cooling) and introduces a single point of failure, making them unsuitable for the majority of remote site implementations.
The next installment in the Evolution of Storage series takes a look at server consolidation and shared SAN/NAS storage. Subscribe to the StorMagic blog and never miss another article again.