Edge Storage for Twelve Factories with Unreliable WAN
A manufacturer operates 12 factories, each with two HPE ProLiant hosts running VMware and a latency-sensitive MES application plus 5 TB of local file shares. Each site has 1 GbE iSCSI and no Fibre Channel. Capacity grows about 20% per year, and the central IT team of four people cannot visit sites for routine changes. The MPLS WAN provides 50 Mbps per factory but has had 8-hour outages twice this year. Factory management requires local operations to continue through a WAN outage, with RPO of 1 hour and local RTO under 30 minutes. Corporate retention requires 30 days of central copies for recovery and audit. The company wants central monitoring and a repeatable per-site build, but the CFO has capped the per-site storage budget and rejected a proposal for a full all-flash data-center array at each factory. The architect must choose between HPE MSA Gen6, HPE Alletra 5000, and HPE Alletra Storage MP B10000, and design replication and management.
1. Select the per-site platform and justify against the other two. Include the minimum HA and capacity configuration.
Read the worked answer
Recommendation: HPE MSA Gen6 at each factory, configured as a dual-controller all-flash or hybrid system with SSD capacity for MES. MSA Gen6 fits the constraints: 1 GbE iSCSI, no FC, snapshots, asynchronous replication, and a lower per-site cost. Alletra 5000 would give stronger analytics and performance, but it is harder to justify under a capped per-site budget for 5 TB of edge data. Alletra Storage MP B10000 is a data-center scale-out platform and is over-designed for a two-host factory with 1 GbE. Minimum configuration: two controllers, dual power supplies, RAID 6 or RAID 10 for MES, two independent 1 GbE switches, multipath on both hosts, and local snapshots every hour. Size from 5 TB * 1.2^3 = 8.64 TB; add 30% free and snapshot reserve, so plan about 12-14 TB usable per site. Use SSD for MES and file metadata; if cost forces hybrid, pin MES volumes to SSD. The condition that changes the recommendation is a requirement for sub-millisecond MES latency or central AIOps across all sites at any cost; then Alletra 5000 becomes defensible.
2. Design replication and central retention over 50 Mbps per site. Show whether RPO 1 hour is feasible.
Read the worked answer
RPO feasibility depends on change rate, not capacity. If each site changes 5% of 5 TB per day, that is 250 GB/day, or 10.4 GB/hour. At 8 bits per byte, 10.4 GB/hour = 23 Mbps; with 2x overhead, about 46 Mbps. That just fits a 50 Mbps link, with little room for other traffic. If the real change rate is 20% per day, it becomes 1 TB/day = 93 Mbps before overhead, which does not fit. Design: take hourly local snapshots and use MSA Remote Snap to replicate them asynchronously to a central MSA for the latest copy; then run a central backup to HPE StoreOnce for 30-day retention. Use compression and dedupe on StoreOnce; schedule replication with QoS so MES traffic is not starved. Initial seeding over 50 Mbps would take about 9 days for 5 TB (5,000 GB * 8 / 50 Mbps), so seed locally at the central site or use a temporary link. For 30-day retention, replicate to a central StoreOnce and keep 30 daily copies; do not try to keep 30 days only on the edge MSA. Test a WAN-outage scenario: local MES must continue from local storage, and replication must catch up automatically. If measured change exceeds the link budget, either relax RPO to 4 hours or upgrade the WAN.
3. How do you deliver central monitoring and repeatable deployment with four IT staff? What do you standardize and what do you monitor?
Read the worked answer
Standardize the build, not the sites. Create a gold configuration: same MSA model, firmware baseline, RAID layout, volume naming, snapshot schedule, replication policy, host multipath settings, and monitoring template. Deploy from a documented runbook and script the host and array settings where possible. With four central staff, remote hands should only rack, cable, and power; configuration should be remote. Monitoring: use the MSA Storage Management Utility for local health and SNMP/email traps to a central collector; where the installed MSA Gen6 firmware and entitlement support HPE InfoSight or HPE GreenLake telemetry, use it for predictive alerts and a single dashboard. Otherwise, standardize on a central monitoring tool that ingests SNMP and syslog. Alert on controller, disk, power, fan, capacity above 70%, snapshot age, replication lag, and failed login. Do not wait for users to report an outage. Operational controls: monthly health checks, quarterly restore tests, firmware review, and a per-site checklist. The goal is a repeatable 30-minute deployment and a known state, not a unique configuration at each factory.