Scality solves data storage challenges: security, performance, and cost. Get unbreakable ransomware protection and help to execute AI data-driven ideas quicker.
SF ☁︎ Paris ☁︎ Tokyo ☁︎ DC
Joined May 2010
- Tweets11.5K
- Following5.2K
- Followers7K
- Likes24.7K
Pinned Tweet
Yesterday we hosted our 11th annual #ScalityDay in Paris, our biggest turnout yet. 🎉 Data, infrastructure and AI in production, cyber resilience, and regulation & sovereignty, capped off by the legendary Scality Rocks after-party. 🎸 Thank you to everyone who joined us!
1
1
146
Official guidance from CISA, the FBI, HHS, USDA and DOT: cisa.gov/resources-tools/res…
Full breakdown: artesca.scality.com/en/blog/…
#cybersecurity #OTsecurity
1
52
Production AI puts new demands on storage. Candida Valois covers what they are and what infrastructure teams should plan for. Join us at #SC26.
📅 Tue Nov 17, 10:30am CT
📍 Room W194b, McCormick Place
Then find us at booth #4959.
sc26.supercomputing.org #HPC #AIInfrastructure
Analyst reports shape how tech gets bought. But how are the rankings actually made? Giorgio Regni and Paul Speciale on how Gartner, IDC, and GigaOm rank storage vendors, what happens when vendors don't participate, and whether AI is changing research: eu1.hubs.ly/H0ygC490
1
85
From GPU cache to object storage to tape: how tiered architectures keep AI data pipelines flowing. Giorgio Regni and Paul Speciale share takeaways from the Supermicro Open Storage Summit, alongside NVIDIA, Intel, and Supermicro: eu1.hubs.ly/H0ygC520
1
102
Kubernetes just got easier for object storage. The COSI driver automates S3 bucket provisioning, IAM credential lifecycle, and cleanup, with built-in observability for SREs. Open source on GitHub. Giorgio Regni and Paul Speciale explain: eu1.hubs.ly/H0ygC500
2
1
113
Enterprise flash prices are up 42%. Is your storage strategy exposed? Answer three questions and get a 0 to 100 storage risk score, see how your flash-to-HDD ratio compares with other Scality customers, and download your report: eu1.hubs.ly/H0ygD4s0
60
100 TB drives are coming. Massive capacity, massive design implications. Giorgio Regni and Paul Speciale on rebuild times, local parity, and the data placement algorithms that keep data durable across drive, server, and rack failures at that scale: eu1.hubs.ly/H0ygC440
128
Everyone's bragging about blazing-fast storage numbers. How real are they? Giorgio Regni and Paul Speciale on why benchmark claims fall apart in production, the hidden cost of accelerator nodes, and why restore speed matters as much as write speed: eu1.hubs.ly/H0ygC510
70
SEC, FINRA, SOX, FFIEC: your storage has to answer to all of them. When regulators look closer, will it hold up? A cyber-resilience checklist mapping storage controls to the regulations financial services IT teams get examined on: eu1.hubs.ly/H0ygD150
63
NAND flash prices expected to surge up to 200%, with supply constraints stretching into 2027. "All-flash everything" is getting hard to justify. Giorgio Regni and Paul Speciale on matching workloads to flash, HDD, and tape, plus a TCO calculator demo: eu1.hubs.ly/H0ygC420
91
"Can object storage do files?" According to Scality CTO Giorgio Regni, that's the wrong question. A decade of lessons on file gateways, where they break, why POSIX and object semantics differ, and what storage architects should ask first: eu1.hubs.ly/H0ygC350
86
Backup is not recovery. Most organizations believe they can recover from ransomware; few have tested it. Our readiness checklist covers the gaps in data, access and dependencies, and how to test restores under real attack conditions: eu1.hubs.ly/H0ygC-l0
2
75
GPUs have staggering compute. The real bottleneck is memory. Giorgio Regni and Paul Speciale explain how KV cache works, why paged attention matters, and how cached blocks tier from HBM to RAM to flash to object storage. Calculator included: eu1.hubs.ly/H0ygC330
1
1
134
October is Cybersecurity Awareness Month. A good time to revisit why ransomware rewrote the rules of backup, and how object lock, checksumming, and layered security make restores faster and safer. Giorgio Regni and Paul Speciale explain: eu1.hubs.ly/H0ygC300
93
Model loading from a single local NVMe drive: ~18 seconds. From distributed object storage over S3: ~12 seconds. Over a native RDMA path: ~2 seconds, close to saturating a 400Gb network. Giorgio Regni and Paul Speciale walk through the test: eu1.hubs.ly/H0ygC400
156
43% of the data hit in a serious incident is never recovered. New research on what actually drives recovery time for mid-size businesses, and why the backup system, not the outage, is usually the limiting factor. Get the report: eu1.hubs.ly/H0yBpzs0
84
Hot in GPU memory, warm in flash, cold on disk, frozen on tape. The storage pyramid assumes data cools over time. With KV caches, long sessions, and RAG constantly refreshing context, does AI data ever go cold? Giorgio Regni and Paul Speciale say no: eu1.hubs.ly/H0yBqsn0
1
126
Is QLC flash the sweet spot for AI storage? Giorgio Regni and Paul Speciale unpack new testing with Samsung at its MAX Lab: 61 TB QLC SSDs, a 400Gb network, 1.46x write amplification, and why $/TB alone is the wrong AI storage metric. eu1.hubs.ly/H0ygBKk0
138