Hewlett Packard Enterprise (HPE) recently launched the latest iteration of its GreenLake for Block Storage solution, positioning it as a revolutionary force in the hybrid cloud AI-based storage market. However, a closer examination reveals potential shortcomings that may render this solution less disruptive than HPE portrays, particularly for enterprise customers and federal government departments.
HPE’s core arguments for GreenLake’s disruptive potential centre on three key features: disaggregated, scale-out block storage, software-defined storage (SDS), and tight integration with AWS. In this review of the technology outlined in their announcement, I will outline key considerations organisations should keep in mind as they review each of the announced offering aspects, for business and technical suitability and value, and explore why they may not be quite the game-changers HPE suggests.
Disaggregation and Scalability: Not a One-Size-Fits-All Solution
- Disaggregation – A Familiar Concept: HPE emphasises its disaggregated architecture, where storage hardware is separated from the software that manages it (typically Fibre Channel or NVMe controllers). This allows independent scaling of compute resources (CPU, memory) and storage capacity (disks, flash). However, disaggregation is not a recent innovation. Established storage vendors like Dell EMC (VxFlex), NetApp (ONTAP with StorageGRID), and Pure Storage (FlashArray//X) already offer similar disaggregated solutions.
- Complexity for IT Teams: While disaggregation offers some flexibility, managing separate hardware and software components can introduce complexity for IT teams. They’ll need expertise in both storage hardware (performance characteristics, troubleshooting) and software (configuration, automation) to ensure optimal performance and troubleshoot issues effectively. Additionally, traditional block storage workloads often rely on specific features from storage controllers, like Quality of Service (QoS) for mission-critical applications. Disaggregation may require replicating this functionality in software, adding further complexity.
- Limited Scalability Needs for Block Storage: The true need for on-demand scalability for block storage is debatable for most enterprises. Block storage is typically used for mission-critical applications requiring predictable performance (low latency, high IOPS) for databases, virtual machines, and high-performance computing. Frequent scaling up or down of storage resources could disrupt application performance and introduce significant management overhead. Additionally, traditional block storage utilises synchronous replication for data mirroring, which can become complex and expensive to manage at scale across geographically dispersed locations.
AI Ops: Hype vs. Reality in Block Storage Management
- Limited Effectiveness in Specialised Environments: HPE emphasises the use of Artificial Intelligence for storage operations (AI Ops) within GreenLake. AI Ops promises automation and optimisation of storage tasks, such as provisioning, tiering, and performance tuning. However, its effectiveness in block storage remains to be seen. Block storage environments are often highly customised for specific workloads with unique performance requirements. AI algorithms may struggle to learn and adapt to these nuances effectively, potentially leading to suboptimal configurations that compromise performance.
- Human Expertise Remains Essential: The concept of AI Ops is still evolving. While it holds promise for automating routine tasks like storage provisioning and basic performance monitoring, it’s unlikely to revolutionise core storage management principles. Experienced IT professionals will still be needed to make critical decisions like workload placement, storage tiering strategies, and configuring advanced features like replication and snapshots. Furthermore, AI may not be able to handle complex troubleshooting scenarios requiring deep storage knowledge.
Cloud Integration: A Double-Edged Sword with Potential Lock-In
- Vendor Lock-In Concerns: The integration of GreenLake with AWS is intriguing, allowing for seamless management of block storage across on-premises and AWS environments. However, this integration raises concerns about vendor lock-in. Customers who heavily invest in GreenLake with AWS may find it difficult and expensive to migrate to a different cloud provider in the future. This lock-in can stem from several factors:
- Proprietary APIs: GreenLake might utilise proprietary APIs to interact with AWS storage services. These custom APIs could differ from the standard AWS APIs, making it challenging to migrate data and workloads to another cloud provider that doesn’t support them.
- Custom Management Tools: HPE might develop custom management tools specifically designed for managing GreenLake within the AWS environment. These tools might not be compatible with other cloud providers, forcing businesses to adopt new tools and processes if they decide to switch cloud platforms.
- Data Lock-In: Migrating large datasets out of the AWS cloud can be time-consuming and expensive due to egress fees. This can create a disincentive for businesses to move away from AWS, even if they find a more cost-effective or functionally superior cloud storage solution from another vendor.
- Public Cloud Not Ideal for All Block Storage Needs: The suitability of a public cloud like AWS for all block storage needs is questionable. Here’s why:
- Latency Concerns: Latency can be an issue for geographically dispersed deployments or applications requiring real-time responsiveness. For instance, applications requiring microsecond response times (e.g., high-frequency trading) may not be suitable for the public cloud due to inherent network latency between the application and the storage resources in AWS.
- Unpredictable Public Cloud Costs: Public cloud storage costs can be unpredictable and may not align well with the fixed-cost budgeting preferred by many government agencies. Costs can fluctuate based on data access patterns, IOPS consumed (number of input/output operations per second), and egress fees (charges for transferring data out of the cloud). Businesses may find it difficult to accurately forecast storage expenses in a public cloud environment.
By carefully considering these factors, businesses need to determine if the potential benefits of cloud integration outweigh the drawbacks. For latency-sensitive applications or those with predictable storage requirements, on-premises block storage solutions may still be the most suitable option.
However, for businesses with hybrid cloud environments or those requiring occasional cloud bursting, GreenLake’s integration with AWS could offer some advantages, but with the caveat of potential vendor lock-in.
The Disaggregated Stack: A Security Challenge
- Disaggregated Security Management: The disaggregated architecture touted by HPE can introduce security complexities. Traditional block storage typically leverages hardware-based security features like secure boot and tamper-evident seals on storage controllers. In a disaggregated environment, these functionalities may need to be replicated in software, increasing the attack surface and reliance on proper configuration.
- Securing Communication Paths: With separate storage hardware and software components, securing communication paths between them becomes critical. Strong encryption protocols like AES-256 are essential to protect data in transit between storage hardware and storage software. Additionally, role-based access controls (RBAC) need to be meticulously implemented to ensure only authorised users can access and manage storage resources.
- Network Security Considerations: The network fabric connecting storage hardware and software becomes a critical security zone. Implementing network segmentation can limit the blast radius in case of a security breach. Additionally, security protocols like IPSec can be used to encrypt communication across the network.
Existing Solutions Can Suffice for Most Needs
- Proven Functionality of Existing Solutions: Existing block storage solutions, when properly implemented and managed, can effectively serve most business needs for the foreseeable future. These solutions offer established functionality with well-understood performance characteristics and a mature support ecosystem. Many existing block storage solutions already offer features like tiering for optimising storage utilisation, snapshots for data protection, and replication for disaster recovery.
- Focus on Optimisation Over Revolution: For many businesses, the focus should be on optimising existing block storage solutions rather than chasing revolutionary new features. This could involve implementing storage tiering strategies to move less frequently accessed data to lower-cost storage tiers like SATA or object storage. Additionally, data deduplication techniques can significantly reduce storage consumption by eliminating redundant data copies. Finally, efficient storage provisioning practices can help avoid storage sprawl and optimise resource allocation. These established best practices can significantly improve storage utilisation and reduce costs without the risks associated with deploying a new and unproven technology.
Wait and Observe Real-World Performance
Given the potential drawbacks mentioned above, a wait-and-see approach may be commercially and technically prudent for organisations considering GreenLake for Block Storage.
- Monitor Early Adopters’ Experiences: Enterprises and government agencies should closely monitor the experiences of early adopters who deploy GreenLake. This will provide valuable insights into the real-world performance, manageability, and security posture of the solution, particularly in terms of how effectively AI Ops handles workload optimisation and how the disaggregated architecture impacts overall operational complexity.
- Evaluate Long-Term Reliability: Assessing the long-term reliability and performance of GreenLake is crucial. Block storage solutions are mission-critical for many organisations, and any potential for instability or performance degradation could have significant consequences. Evaluating reliability metrics like uptime and mean time to repair (MTTR) from early adopters will be crucial before widespread adoption.
Evaluating GreenLake Suitability For Specific Needs
HPE’s GreenLake for Block Storage represents an interesting foray into the hybrid cloud storage market. However, its disruptive potential seems exaggerated. Established storage solutions offer proven functionality, and the benefits of disaggregation, AI Ops, and public cloud integration may not be as significant as HPE suggests.
Businesses, particularly federal government departments, should carefully consider their specific storage requirements and weigh the potential advantages of GreenLake against the risks and uncertainties associated with a new and unproven technology.
A measured approach, focusing on optimising existing solutions and closely monitoring early adopter experiences, will allow businesses to make informed decisions about whether or not GreenLake aligns with their long-term storage needs.
Businesses should also consider the technical complexities associated with the disaggregated architecture, particularly regarding security implications and the additional burden placed on IT staff for managing separate hardware and software components.
