HPE ASE Storage Architect flashcards

32 cards. Read the term and the definition here. The deck at the top of this page is optional.

Usable capacity formula

Usable capacity equals raw capacity multiplied by one minus parity overhead fraction, then multiplied by one minus reserved spare percentage. RAID 6 across eight data disks loses 25 percent to parity.

RAID level selection rule

Use RAID 10 for write-heavy low-latency workloads, RAID 5 for read-heavy sequential access, and RAID 6 when large drives make rebuild risk unacceptable. RAID 0 never protects data.

IOPS per drive class

Approximate random IOPS: NVMe SSD tens to hundreds of thousands, SAS SSD around ten thousand or more, 15K SAS HDD about 180, and 7.2K NL-SAS about 75.

Throughput from IOPS

Sequential throughput in megabytes per second approximately equals IOPS multiplied by block size in kilobytes, divided by 1024. Validate both IOPS and bandwidth against the workload profile.

Thin provisioning overcommit ratio

Overcommit ratio equals provisioned capacity divided by physical capacity. Keep it below roughly 2:1 unless pool monitoring alerts fire at 70 to 80 percent consumption.

Erasure coding overhead

With k data fragments and m parity fragments, the scheme tolerates m simultaneous failures and incurs overhead of m divided by k. Eight data plus two parity equals 25 percent overhead.

Latency tier decision rule

Application SLA below 1 millisecond requires all-flash NVMe. One to five milliseconds fits hybrid SSD tiers. Above 10 milliseconds tolerates HDD-backed capacity tiers.

Protocol selection by workload

Fibre Channel suits deterministic low-latency block on SAN fabrics, iSCSI suits cost-sensitive block over Ethernet, NFS and SMB suit file sharing, and NVMe-oF targets sub-100-microsecond block.

Single-initiator FC zoning

Zone by worldwide port name with one host HBA port and one array target port per zone. Never zone host-to-array traffic by worldwide node name, and keep each HBA on one fabric.

ALUA path states

Asymmetric logical unit access reports paths as active-optimized or active-non-optimized. Multipath software sends I/O through optimized paths and uses non-optimized paths only for failover.

iSCSI deployment practices

Place storage traffic on a dedicated VLAN, enable jumbo frames end-to-end or not at all, authenticate with CHAP, and configure MPIO with at least two independent sessions.

LUN masking sequence

Create the host object with its WWPN or IQN, create the volume, map the volume to the host or host set, then rescan the host bus and verify paths with OS multipath tools.

Array deployment sequence

Install hardware, assign management addressing, apply firmware baseline, build pools or RAID groups, create volumes, map to hosts, configure multipath, create filesystems, then validate the application.

NVMe-oF fabric requirement

NVMe over Fabrics needs a lossless or congestion-controlled network, such as priority flow control and ECN for RoCEv2, plus discovery controller configuration and a host namespace scan after connect.

Interoperability matrix rule

Array controller firmware, drive firmware, host bus adapter firmware, and multipath driver versions must all appear together in the vendor interoperability matrix before production cutover.

Snapshot reserve sizing

Reserve capacity equals expected change rate multiplied by the retention window. Monitor snapshot pool consumption so the array does not force deletion of the oldest snapshots.

Migration cutover steps

Baseline performance, replicate data, quiesce the application, perform final synchronization, remap host paths, validate application behavior, then decommission the source after the rollback window.

Performance troubleshooting order

Check host queue depth and multipath path state first, then fabric or network errors, then array front-end port utilization, and finally backend disk latency and saturation.

Utilization escalation thresholds

Sustained array port utilization above 70 percent, backend disk busy above 80 percent, or pool consumption above 80 percent are typical triggers for escalation and remediation.

Host queue depth formula

Optimal queue depth approximates IOPS per path multiplied by average latency in seconds. Setting depth far above this raises latency without increasing delivered throughput.

Alert tiering design

Separate predictive warnings from critical component failures, and deduplicate alerts by managed object so a single failed disk does not flood the ticketing system.

Capacity trend reporting

Report subscribed, used, and snapshot overhead separately. Thin pool growth rate trends predict the exhaustion date more usefully than a single current utilization percentage.

Change management requirements

Every firmware or zoning change needs a documented rollback plan, a scheduled maintenance window, and post-change validation of multipath state and application latency.

Replication lag monitoring

Monitor recovery point objective lag, not only link state. A replication link that reports up but lags silently breaches the recovery point objective.

Log collection for escalation

Capture array event logs, host multipath logs, and fabric logs with synchronized NTP timestamps before escalation so events can be correlated across layers.

RPO versus RTO

Recovery point objective is the maximum tolerable data loss; recovery time objective is the maximum tolerable downtime. Synchronous replication gives near-zero RPO; asynchronous gives RPO equal to the interval.

Snapshot is not backup

A snapshot shares the source volume failure domain and does not protect against array loss or corruption. A backup requires an independent copy on separate media or site.

Immutable copy defense

Immutable or air-gapped copies with retention lock cannot be modified or deleted for a defined period, defending backups against ransomware encryption and insider deletion.

Encryption scope limitation

Data-at-rest encryption protects against physical drive theft, not against a compromised host with mounted volumes. Use host or application encryption for that threat.

Key management redundancy

Use an external key manager with at least two key servers. Losing all key copies makes encrypted data unrecoverable, including every backup copy of that data.

Application-consistent replication

Crash-consistent copies reflect an abrupt power loss. Application-consistent copies quiesce the application or use VSS snapshots before replication so recovery needs no repair.

RBAC and audit logging

Separate array administrator, storage operator, and read-only roles. Enable audit logging of configuration changes and forward those logs off-array to protect them from tampering.