Consolidating a 3PAR and Nimble Estate for a Logistics Firm
A regional logistics company runs 180 VMware VMs on an aging HPE 3PAR 8200 and an HPE Nimble CS300 at its primary site. The DR site is 40 km away and currently has only tape and snapshot copies. Business services include Oracle order management, SQL Server warehouse, and file services. Current usable block capacity is 40 TB, with 25% annual growth forecast for three years. The CIO wants to retire both arrays within two quarters, keep RPO at 15 minutes or better and RTO under 4 hours, and add cloud-based monitoring because the storage team has three people. The network team can provide 10 GbE for replication but the hosts use 8 Gb FC with two independent fabrics. Finance prefers an operating-expense model but will consider capex if the five-year cost is clearly lower. The customer is unsure whether HPE MSA Gen6, HPE Alletra 9000, or HPE Alletra Storage MP B10000 is the right fit and asks for a defensible recommendation.
1. Recommend the primary and DR platform, and justify the choice against MSA Gen6 and Alletra 9000. Include the one condition that would change your recommendation.
Read the worked answer
Recommendation: HPE Alletra Storage MP B10000 at both sites. The decisive requirements are 180 VMs, Oracle and SQL, 25% growth, RPO 15 minutes, and a three-person team. MSA Gen6 is an entry platform: dual controllers, iSCSI/FC, snapshots, and async replication are useful, but it is not the best fit for a consolidated Oracle plus VMware estate with multi-node resiliency and scale-out growth. Alletra 9000 is a mission-critical all-NVMe platform with Peer Persistence and synchronous replication; it is technically capable but economically over-specified because the stated RPO is 15 minutes, not zero. Alletra Storage MP B10000 gives disaggregated scale-out block storage, multi-node resilience, NVMe performance, and cloud-based AIOps through HPE InfoSight/GreenLake. Use asynchronous replication with 15-minute snapshots. The condition that changes the answer is a requirement for zero RPO or transparent failover for Oracle: then Alletra 9000 with Peer Persistence and a third-site quorum witness becomes the stronger option, assuming the added cost and network latency are acceptable. Also verify the exact replication and witness support in the current B10000 firmware and HPE sizing tool before committing.
2. Size usable and raw capacity for three years, and state a defensible replication bandwidth estimate. Show assumptions.
Read the worked answer
Capacity math. 40 TB * 1.25^3 = 78.1 TB of active data after three years. Do not fill an array to 100%. If the design target is 70% maximum utilization and you reserve 15% for snapshots and replication metadata, plan roughly 78.1 / 0.70 = 112 TB usable, then add snapshot reserve: 112 * 1.15 = 129 TB usable. Raw capacity depends on RAID and sparing; use HPE sizing tools rather than a hand calculation, but a rough 150-170 TB raw range is defensible. Verify dedupe/compression claims against the actual Oracle and SQL data because database redo and some file data dedupe poorly. Replication bandwidth. Assume 5-10% of active data changes per day. At 10% of 78 TB, daily change is 7.8 TB; a 15-minute RPO means 7.8 TB / 96 = 81 GB per interval, or about 720 Mbps average (81 GB * 8 / 900 s). Add 2x for protocol and burst overhead: about 1.4 Gbps. A 10 GbE link is therefore appropriate; 1 GbE would be marginal or inadequate. If change is only 5%, the average halves to about 360 Mbps and 10 GbE still gives headroom. Seed the DR copy locally or with a temporary high-bandwidth link, then throttle and monitor.
3. The CFO asks for a GreenLake subscription vs capex comparison. What financial and technical factors decide, and what risks do you flag?
Read the worked answer
GreenLake vs capex is not just a financing question. Choose GreenLake when the customer values cash preservation, a 3-5 year subscription with capacity buffer, built-in monitoring, and predictable refresh; it also helps if growth is uncertain and the customer wants to avoid overprovisioning. Choose capex when the five-year discounted cost is clearly lower, the workload is stable, and the customer has capital and depreciation capacity. Build the comparison on the same scope: two sites, replication, support, software, installation, power/cooling, staff time, and refresh. Ask for the minimum commit, overage rate, metering method, early-termination terms, data-sovereignty constraints, and what happens at contract end. Flag risks: a 25% growth forecast can be wrong; GreenLake minimums may force payment for unused capacity; a DR-site subscription may be underused; network egress and cloud monitoring dependencies need review. Recommend a 3-year GreenLake quote against a 5-year capex TCO if the CIO wants opex, but state that capex wins if the modeled five-year cost is more than 15-20% lower and growth is stable. Do not present the subscription as always cheaper.