As customers grow their storage footprint, they typically must add staff to manage the larger environment.  The challenge is to remove much of the complexity that goes with managing modern storage systems.  A new system must perform de-clustered rebuilds and take care of replacing lost resources without requiring human intervention.  Users need to replace isolated islands of storage with a single system that scales.  Scalable fault tolerance, built-in automation, hardware monitoring, and a simple management interface greatly diminish the need for this approach.

To find out more, VMblog spoke with industry expert Björn Kolbeck, co-founder and CEO Quobyte.

VMblog:  What kind of system best delivers the high performance and low latency results required of organizations today?

Björn Kolbeck:  A POSIX-compatible file system that delivers remarkable scale-out performance without driving up headcount or maintenance budgets ensures that customers will never outgrow their storage solution.  The system should offer true linear scaling - double the number of storage nodes and attain double the performance.  Unified storage allows multiple users to work simultaneously on the same file regardless of the access protocol being used.  And the system should be able to be monitored, maintained and healed with a small team to allow for capacity growth from terabytes to petabytes while still delivering access to data despite hardware failures.

VMblog:  Why does block storage fail in scale-out storage scenarios?

Kolbeck:  Tracking millions or billions of storage blocks adds complexity, hampers resizing operations, requires a large memory footprint, and increases inter-node cache synchronization headaches.

VMblog:  Is object storage a good choice for storage requiring high IOPs performance?

Kolbeck:  While object storage's handling of billions of entities makes sense from a hyperscale perspective, it was never designed for high IOPs, especially in small-file workloads.  Applications would either need to be rewritten to take advantage of objects completely different set of access semantics or a gateway would have to be put in place to provide file-like semantics.  And while object storage's eventual consistency model is adequate for archival, it doesn't meet the needs of primary workloads that expect a POSIX-compatible file system.

VMblog:  How do I ensure my storage system doesn't suffer from performance issues and/or capacity bottlenecks?

Kolbeck:  Understanding and predicting storage needs can be difficult, especially with multiple workloads and users on the system at the same time.  Built-in real-time analytics let administrators pinpoint the highest IOPs, throughput and metadata consumers while external analysis of long-term trends can be studied to identify potential performance and bottleneck issues before they become actual problems.

VMblog:  Why is having broad support and built-in integration of multiple technology platforms critical?

Kolbeck:  Today's storage needs to work in today's ecosystems.  By providing broad support for everything from OpenStack to bare metal, finding a storage solution that can meet an organization's needs regardless of where they are on the technology adoption scale is essential.  A system that offers proxies to support various protocols ensures that applications retain connectivity to the storage cluster for high performance.

VMblog:  What reliability features are needed to keep data safe within a storage system?

Kolbeck:  Keeping data safe and available in the event of hardware or software failures are the most important tasks of any storage system.  A storage system needs to address data safety and integrity through multiple layers of protection, such as replication, erasure coding, failure domains, checksums, monitoring and automated health managers.  These services can make all the difference in keeping the storage cluster operating properly, maximizing resource utilization and eliminating potential failures.

VMblog:  If something catastrophic does happen to my storage system, what tools can be deployed to ensure recovery from outages?

Kolbeck:  Whether from natural or manmade causes, the ability to withstand a catastrophic outage must be factored into a company's disaster recovery planning.  Volume mirroring gives administrators the ability to remotely mirror a volume from one cluster to another, regardless of location.  Continuous asynchronous replication can be done on a per-file basis.  Source volumes can be mirrored to multiple target clusters for even greater levels of redundancy.

VMblog:  How can companies continue to manage a growing storage environment without having to add staff?

Kolbeck:  Quobyte software provides non-stop, high-performance storage for the most- demanding workloads using economical, commodity hardware and user-friendly monitoring tools. The solution's linear scaling, run-time configuration flexibility and real-time performance monitoring let organizations respond to changing storage requirements instantly. Support for rolling upgrades and non-disruptive cluster expansion remove the burden of planned outages. Companies can choose how data is stored with data delivered to clients over the broadest set of access protocols, all while maintaining consistent access control. Finally, Quobyte delivers storage freedom by enabling customers to combine the benefits of flash with the storage efficiency of hard drives, never forcing them to use one when the other will do.

##

Before taking over the helm at Quobyte, Björn Kolbeck spent time at Google working as tech lead for the hotel finder project (2011-2013). He was the lead developer for the open-source file system XtreemFS (2006-2011). Björn's PhD thesis dealt with fault-tolerant replication.