PowerStore Architecture and Hardware Components
This domain tests whether you can describe what is inside a PowerStore appliance, how appliances combine into a cluster, and where the T and X families diverge. Expect questions on node count, media type, I/O modules, and scale limits rather than on marketing speeds.
Appliance anatomy
- Every PowerStore appliance contains two active-active nodes (node A and node B). Each node has its own CPU, memory, NVRAM used to persist writes before they land on media, and a dedicated management port.
- Media is NVMe SSD only: 2.5-inch NVMe drives inside the appliance, plus external NVMe expansion enclosures when more capacity is needed. There are no SAS or NL-SAS drives, so any answer that describes tiering between SAS and SSD is wrong for this platform.
- RAID is built automatically from the NVMe drives; you do not hand-build RAID groups or assign hot spares manually.
- Each node has I/O slots for front-end modules: 4-port 25 GbE SFP28, 4-port 32 Gb FC, and higher-speed options on larger models. Those ports carry host I/O and, on some configurations, inter-appliance traffic.
- PowerStore T models serve both block (volumes over FC, iSCSI, NVMe-oF) and file (SMB/NFS through NAS servers), so a single appliance can present both to the same environment.
Scale up and scale out
- Adding drives or an expansion enclosure scales an appliance up.
- Adding appliances scales the cluster out. One PowerStore cluster supports up to four appliances, managed as a single system from one PowerStore Manager session.
- Volumes remain owned by the appliance that hosts them; when a new appliance joins, you decide which volumes live where rather than migrating everything.
- PowerStore X models run an embedded hypervisor (AppsON) so virtual machines can run on the array itself; PowerStore T is the unified block-and-file model most sites deploy.
Worked example
A customer needs roughly 300 TB usable now and warns that a second system will be bought next year. Reasoning: size a T model with internal NVMe drives plus expansion enclosures, confirm the count stays inside the four-appliance cluster ceiling, and have the second unit purchased as a cluster member rather than a standalone array. That way it appears under one management IP, one alert stream, and one capacity view instead of two silos.
Traps
- Claiming PowerStore uses SAS SSDs, or that NVRAM is a read cache rather than a write-persistence buffer.
- Assuming hosts must be reconfigured when an appliance joins the cluster.
- Treating PowerStore X and T as interchangeable when the question specifies AppsON or embedded ESXi.