As 2012 draws to a close, it’s a good time to look forward. Storage networking is evolving rapidly, driven by virtualization, cloud, flash, and scale requirements. Based on what I’m seeing at Cisco and in the broader industry, here are my predictions for where storage networking is heading.
Flash Becomes Mainstream
Flash is transitioning from exotic to standard:
All-Flash Arrays: Within 2-3 years, all-flash arrays will be common for Tier 1 workloads. Costs are dropping rapidly, making flash economically viable for more workloads.
Hybrid Becomes Standard: Even commodity arrays will include SSD for caching or tiering. Pure spinning-disk arrays will be relegated to archives and backup.
NVMe Protocol: As flash becomes standard, we’ll see protocols optimized for flash. NVMe (NVM Express) will replace SATA/SAS for SSDs, providing much lower latency.
Storage Class Memory: Technologies like 3D XPoint will blur the line between memory and storage, enabling new architectures.
Flash will fundamentally change storage performance expectations. What we consider great performance today will be merely adequate tomorrow.
Software-Defined Storage Grows
The software-defined trend will reshape storage:
Decoupling: Storage software will increasingly decouple from hardware, running on commodity servers.
Scale-Out: Software-defined storage will enable massive scale-out on commodity hardware.
Open Source: Projects like Ceph and OpenStack Swift will mature and gain enterprise adoption.
Hybrid Clouds: Software-defined storage will enable seamless integration between on-premises and cloud storage.
This doesn’t mean the death of purpose-built arrays—they’ll still have a place for certain workloads—but we’ll see software-defined storage grow significantly.
Convergence Accelerates
FCoE is just the beginning:
Unified Fabric: Ethernet will carry more protocols—storage, data, management. 10 GbE will become standard, 40 GbE will emerge.
Hyper-Convergence: Compute, storage, and networking integrated into single appliances. VMware VSAN and Nutanix are early examples.
Simplified Operations: Fewer separate networks means simpler operations and lower costs.
The trend is clear: collapsing multiple networks onto unified infrastructure. This requires careful design but offers significant benefits.
Cloud Integration Deepens
Cloud won’t replace on-premises storage, but integration will deepen:
Hybrid Cloud: Seamless data movement between on-premises and cloud.
Cloud Tiering: Automatic movement of cold data to cloud storage.
Cloud Bursting: Using cloud capacity for temporary needs.
Disaster Recovery: Cloud as DR target becomes common.
Backup to Cloud: Cloud replacing tape for long-term retention.
Most enterprises will run hybrid environments, with storage spanning on-premises and cloud.
Automation and Orchestration
Manual storage management won’t scale:
Policy-Based Management: Define policies, let software handle implementation.
Self-Service: Developers and applications provision storage via APIs without admin intervention.
Automated Tiering: Continuous optimization of data placement across tiers.
Predictive Analytics: Machine learning predicts failures and optimizes placement.
Infrastructure as Code: Storage infrastructure defined in code, version controlled, automatically deployed.
The future storage admin is more programmer than operator.
Object Storage Growth
Object storage will grow beyond cloud:
Scale: Object storage scales to billions of objects, enabling massive unstructured data repositories.
Cost: Optimized for cost per TB, not performance.
Applications: More applications will be built for object storage rather than block/file.
On-Premises: Object storage will deploy on-premises, not just in cloud.
Unified Namespace: Global namespace spanning locations.
Object storage complements rather than replaces block and file storage.
Data Reduction Technologies
Deduplication and compression will become ubiquitous:
Inline: Real-time deduplication and compression, not just post-process.
Higher Efficiency: Better algorithms providing better reduction ratios.
Less Overhead: More efficient implementation with lower performance impact.
Global Deduplication: Dedupe across arrays or even across data centers.
As flash adoption grows, data reduction becomes more important to manage flash’s higher cost per GB.
Network Speeds Increase
Network bandwidth will continue growing:
10 GbE Mainstream: 10 Gigabit Ethernet becomes standard for server connectivity.
40/100 GbE Core: Data center cores move to 40 or 100 Gigabit Ethernet.
FC Speeds: 16 Gbps FC is arriving, 32 Gbps in development.
Low Latency: Focus shifts from just bandwidth to latency. Latency-optimized networks emerge.
Higher bandwidth enables richer storage protocols and convergence.
Storage Virtualization Evolution
Storage virtualization will continue evolving:
Greater Intelligence: More sophisticated optimization algorithms.
Cross-Protocol: Virtualizing across block, file, and object storage.
Cloud Extension: Extending virtualization to include cloud storage.
Application Integration: Tighter integration with application platforms.
Automation: Hands-off management with policy-driven automation.
FC-Redirect is early in this evolution. We’ll see much more sophisticated storage virtualization.
Container Storage
Containers are emerging and need storage:
Persistent Storage: Containers need persistent storage that survives container restarts.
Portability: Storage that moves with containers across hosts and clusters.
Multi-Tenancy: Isolation between containers.
Performance: Low-latency storage for containerized applications.
Container storage is nascent but will grow as container adoption increases.
Data Protection Evolution
Backup and DR continue evolving:
Continuous Data Protection: Moving from periodic backups to continuous protection.
Application-Integrated: Backup more tightly integrated with applications for consistency.
Cloud Backup: Cloud as backup target becomes standard.
Instant Recovery: Recovering VMs directly from backup storage without restore.
Ransomware Protection: Backup isolation to protect against ransomware.
The focus shifts from “backup” to “protection and recovery.”
AI and Machine Learning
AI will optimize storage:
Predictive Failures: Predicting drive failures before they occur.
Automated Tiering: ML algorithms optimizing data placement.
Capacity Forecasting: Better prediction of capacity needs.
Anomaly Detection: Detecting unusual patterns that indicate problems.
Performance Optimization: Continuous tuning of parameters for optimal performance.
AI will make storage more self-managing and self-optimizing.
NVMe Over Fabrics
NVMe will extend beyond local connections:
NVMe-oF: NVMe over Fabrics enables networked NVMe storage.
RDMA: Remote Direct Memory Access provides ultra-low latency.
Latency: Sub-10 microsecond latency becomes achievable.
New Architectures: Ultra-low latency enables new application architectures.
This could be as significant a shift as FC was from SCSI.
Persistent Memory
Memory and storage boundaries blur:
Non-Volatile Memory: Intel 3D XPoint and similar technologies.
Byte-Addressable: Unlike block storage, directly addressable at byte level.
Performance: Performance between DRAM and flash.
New Paradigms: Applications designed for persistent memory will differ from traditional storage-based applications.
This is further out but could fundamentally change storage architecture.
Security Focus Increases
Storage security will become more critical:
Encryption Everywhere: Data-at-rest encryption becomes standard.
End-to-End Encryption: Data encrypted from creation to destruction.
Key Management: Sophisticated key management systems.
Zero Trust: Security models that trust nothing by default.
Compliance: Increasing regulatory requirements drive security investment.
Security will be designed in from the start, not added later.
Multi-Cloud Management
Managing storage across clouds:
Cloud-Agnostic: Tools that work across AWS, Azure, Google Cloud, etc.
Data Mobility: Moving data between clouds.
Cost Optimization: Choosing optimal cloud for each workload.
Unified Management: Single pane of glass across all clouds.
Multi-cloud complexity drives demand for better management tools.
Edge Computing Storage
Computing is moving to the edge:
Edge Storage: Storage at edge locations close to data generation.
Synchronization: Syncing between edge and core data centers.
Bandwidth Management: Managing limited bandwidth between edge and core.
Local Processing: Processing data locally before sending to cloud/data center.
IoT and edge computing create new storage requirements.
Protocol Evolution
Storage protocols will continue evolving:
FC Continues: FC won’t disappear—it continues evolving for mission-critical workloads.
iSCSI Remains Relevant: Cost-effective alternative for appropriate workloads.
FCoE Niche: FCoE finds its niche in specific use cases.
NVMe-oF Emerges: New protocol for flash-optimized workloads.
Object APIs: S3 API becomes de facto standard for object storage.
We’ll see coexistence of multiple protocols, each optimal for specific use cases.
Sustainability Focus
Environmental concerns drive change:
Energy Efficiency: More efficient storage consuming less power.
Higher Density: More capacity per rack reduces data center footprint.
Flash Adoption: SSDs use less power than spinning disks.
Intelligent Power Management: Powering down idle components.
Green Procurement: Consideration of environmental impact in purchasing decisions.
Sustainability becomes a factor in storage architecture decisions.
Skill Set Evolution
Storage professionals need evolving skills:
Programming: Automation requires coding skills (Python, etc.).
Cloud Platforms: Understanding AWS, Azure, Google Cloud.
Networking: Deeper networking knowledge as storage and networks converge.
Security: Security expertise becomes essential.
Data Analytics: Using analytics for capacity planning and optimization.
The future storage professional is more hybrid—storage, networking, programming, cloud.
Organizational Changes
How organizations manage storage will change:
DevOps: Storage integrated into DevOps workflows.
Infrastructure as Code: Storage infrastructure managed as code.
Cross-Functional Teams: Storage, networking, and server teams merge.
Self-Service: Users provision storage without tickets to storage team.
Outsourcing: More storage consumption via cloud, less internal management.
The traditional storage administrator role will evolve significantly.
What Won’t Change
Amid all this change, some things remain constant:
Reliability: Data must be protected and always accessible.
Performance: Applications will always want faster storage.
Capacity: Data volumes continue growing.
Cost Pressure: Organizations want to do more with less.
Fundamentals: The physics of storage—latency, bandwidth, capacity—don’t change.
Understanding fundamentals remains valuable even as technologies change.
Predictions Summary
My predictions for the next 3-5 years:
- Flash becomes mainstream for most workloads
- Software-defined storage grows significantly
- Cloud integration becomes standard
- Automation and AI optimize storage
- Object storage expands beyond cloud
- NVMe revolutionizes storage performance
- Security becomes built-in, not added on
- Multi-cloud becomes the norm
- Storage and networking further converge
- Skills required for storage professionals evolve significantly
Conclusion
The next few years will see dramatic changes in storage networking. Flash, cloud, software-defined architectures, and automation are reshaping the industry.
For storage professionals, this is exciting but challenging. The skills that made you successful won’t be sufficient. Continuous learning is essential.
The good news: storage remains critical. Despite all the changes, applications need to store and retrieve data reliably and quickly. Those who understand both traditional and emerging storage technologies will be well-positioned.
Working at Cisco on FC-Redirect has given me a front-row seat to this evolution. Storage virtualization is part of this transformation—providing the abstraction layer that enables flexibility and automation.
The future of storage networking is bright. The technologies we’ll have in 5 years will make today’s storage seem quaint. But the fundamental principles—reliability, performance, efficiency—endure.
I’m excited to be part of this evolution. The storage world of 2017 will look very different from 2012. Let’s build that future together.