19 When ransomware hits, every minute of downtime costs you. New performance data from Scality and Commvault shows that scale-out object storage now delivers restore throughput that flash-only backup targets can’t match at any capacity that matters. The cost of a ransomware incident goes well beyond lost revenue. For over 90% of mid-size and large enterprises, a single hour of downtime costs more than $300,000. For 41%, it ranges from $1 million to over $5 million. And that’s before litigation, fines, or regulatory penalties1. In the worst cases, businesses don’t reopen. When you’re staring at encrypted production systems, last night’s backup speed is irrelevant. What matters is how fast you can read your data back and put the business online. And regulators have caught up to this math. Under frameworks like DORA and NIS2, the breach is only part of the liability. You must test and demonstrate that operations can be restored inside a defined recovery-time objective. Slow recovery is its own violation. The all-flash backup argument is built on the wrong assumption For years, the backup industry has equated “fast restore” with “all-flash storage.” Vendors like Everpure (FlashBlade//E) and VAST Data have built their backup-tier pitch on the assumption that NVMe TLC and QLC flash equal faster recovery. The argument sounds reasonable until you look at the math. Restore speed at scale isn’t a function of media latency. It’s a function of how many parallel data streams your storage layer can serve to the backup application simultaneously, at the petabyte volumes your real backups occupy. Media speed is a per-drive characteristic. Restore throughput is an aggregate, scale-out characteristic. The two are not the same thing — and at the scale most enterprises operate, they diverge sharply. What the numbers actually show Recent performance testing of Commvault on HPE ProLiant DL345 Gen12 servers demonstrates restore throughput across three Scality configurations: RING all-flash, RING hybrid and ARTESCA. The charts below show effective throughput based on 2× data reduction. Capacities are post EC-usable (RING 9+3, ARTESCA 9+1/5+1), with capacity measured as usable storage after 9+3 erasure coding. Extreme restore performance on all-flash Scality RING. Cost-effective high-performance on hybrid Scality RING. Affordable high-performance with Scality ARTESCA. Scality RING on NVMe flash can be scaled to deliver 7,200 TB per hour of effective restore throughput for Commvault. The same architecture scale-out on hybrid HDD nodes sustains 4,333 TB/hour. For mid-market environments, Scality ARTESCA delivers up to 150 TB/hour at petabyte scale with the operational simplicity Commvault administrators expect from an integrated backup target. To put 7,200 TB/hour in context for Commvault-based restores: at that rate, a 50 TB recovery completes in about 25 seconds. A petabyte restores in under nine minutes. A 10 PB data center comes back in under one and a half hours. Why object storage wins through parallelism and scalability Three architectural reasons explain why this gap isn’t a fluke that flash vendors will close in the next product cycle. Parallel scale-out by design. Scality RING and ARTESCA were built from day one to serve thousands of concurrent S3 connections, which is exactly how Commvault MediaAgents pull data during a multi-node restore. Adding nodes adds throughput close to linearly. There is no controller bottleneck waiting at higher concurrency. Capacity headroom at the same architecture. All-flash targets don’t hit a capacity wall; they hit a cost wall at PB scale. Even before the current media price super-cycle, QLC flash cost roughly 5–7x per usable TB versus the HDD tier in a hybrid object store. Beyond that, customers split into multiple appliances and manage them as separate islands of storage. Scality RING runs the same architecture from 100 TB to 100+ PB with the same Commvault integration, the same performance characteristics, and the same operational model. The Tier 1 hot tier and Tier 2 retention tier live on the same platform. No flash tax where flash doesn’t help. With Scality, you put NVMe exactly where it moves the restore needle — on the Tier 1 primary copy — and you put HDD where capacity economics matter. All-flash backup targets force you to pay flash prices for the 80% of your backup footprint that sits idle for months at a time. Speed without trust is worthless If your backup data has been silently corrupted, exfiltrated, or encrypted before you tried to restore, recovering it fast just gets you back to the breach faster. That’s why Scality pairs performance with CORE5 cyber-resilience in the same product: API-level S3 Object Lock for instant configurable immutability, zero-trust IAM for fine-grained access control, geographic dispersion via multi-site replication and stretched clusters, and architectural immutability that survives even an administrator compromise. Combined with Commvault’s WORM-protected workflows, the result is verified, untouched backups and the raw throughput to restore them at the speed the business needs. Scality and Commvault: Proven in production Scality and Commvault count over 60 joint customers globally, from enterprises to government agencies, managing over 300 PB of Commvault backup data on Scality storage. The performance numbers above aren’t a lab demo. They’re how this architecture behaves under real Commvault workloads at the scale our customers run. If your current backup target makes you choose between fast restore and economical scale, or worse, between immutability and performance, it’s time to revisit the assumption. 1 Information Technology Intelligence Consulting (ITIC), “ITIC 2024 Hourly Cost of Downtime Report,” Parts 1 and 2, September 2024. ITIC reports that the average cost of a single hour of downtime now exceeds $300,000 for over 90% of mid-size and large enterprises, and that 41% of enterprises estimate hourly downtime costs at $1 million to over $5 million. ITIC also notes that these figures exclude litigation, fines, civil penalties, and criminal penalties. (ITIC Corp)
