What Matters in Storage

Barbara Murphy. July 5, 2019
What Matters in Storage

What Matters in Storage

What matters in storage? The past few decades have brought an avalanche of innovations to the storage industry, such as tiering, snapshots, deduplication, replication, iSCSI, SSD, NVMe, and several others, but the core of what storage itself delivers really hasn’t changed. This might sound shocking, but let me explain by first establishing a definition of a storage service.

A storage service is the combined delivery of three interdependent services: capacity, performance (defined by throughput, IOPS, and latency), and availability or up-time. Again, these core services haven’t changed over time, except in one important aspect: the degree to which they need to be delivered in order to meet the demands of today’s applications and workloads.  This doesn’t mean that the innovations we mention above aren’t important; on the contrary, each one is designed to improve the delivery of one or more of those three core storage services.  So, what does drive the adoption of innovation? There is only one answer: cost.

Reducing the cost of delivering a storage service is what has been driving storage innovation. Here are some examples:

  • Deduplication reduces the amount of actual data being stored, thus reducing capacity required to support an application’s needs. If storage capacity costs $1 and deduplication reduces the need by 50%, then even if the cost of deduplication is $0.20, the overall savings of $0.30 would justify adoption. If there were no cost advantages to deduplication, would it matter whether the system had it or not?
  • Tiering is the ability to move data across different media types, making it possible to place the data on storage that is most suitable based on performance needs. Moving data to a lower-performance but less expensive tier, where storage could cost 3X less, would result in significant cost efficiencies.
  • iSCSI as an access protocol that replaced Fibre Channel (FC) in many data centers. Though initially the performance of iSCSI was viewed as lower than that of FC, the cost of deploying an iSCSI network was significantly less expensive. Many applications were fine with iSCSI performance, and the adoption of this protocol skyrocketed because of its extensive cost savings.
  • MLC, TLC, and QLC are solid state disk (SSD) technologies based on NAND* flash that deliver higher density with each iteration. The adoption of higher density NAND offers a lower per Gigabyte (GB) capacity cost. Since NAND delivers better performance per GB than hard disk drives (HDD), the cost savings are amplified with higher density.

There are many other examples where innovations in storage were targeting improvements in performance, capacity density and utilization, up-time, and data access. All of them drive down the overall cost of storage services and help align total cost of ownership with business value, which is what matters in storage.

WekaIO Storage Services

WekaIO has designed its file system to be flexible in delivering cost-effective storage services with the agility to respond to the changing application needs. The two-tier architecture enables the system to deliver performance and capacity independent of each other. Here are some ways in which WekaIO can be customized:

  • The system can be scaled out for greater throughput where all the nodes in the cluster work together in parallel. In addition to scalability, the scale-out design delivers a greater level of resiliency, ensuring that applications have full data access without performance degradation even in the event of a node failure.
  • The WekaIO file system can use any SSD interface (SATA, SAS or NVMe) to increase density or improve performance. If the workload is write-intensive, higher resiliency TLC NAND flash can be used. Conversely, for a read-intensive workload, QLC NAND may provide greater value.
  • Most workloads have active and static data; however, static data doesn’t mean it’s never read. It is typical for applications to access less than 20% of the overall data set at any point in time. Storing all this data in NAND flash can be expensive, even with deduplication and compression. WekaIO can leverage high density HDD via any S3 object storage to store the static data, resulting in a much more cost-effective solution as HDD media continues to have a 5x cost advantage over flash. The benefits of using S3 object storage extend to providing geographic redundancy with less capacity overhead than in traditional storage environments. A geo-dispersed object storage system can deliver site resiliency without the need to create a complete replica of the data set. Compared to a traditional replicated system that typically has a more than 2X ratio of raw to usable, the cost savings are significant.
  • Many users are leveraging on-demand public cloud to augment compute resources for peak work periods. This requires a storage service that can seamlessly integrate with the public cloud. The WekaIO file system is a software-defined stack that can run on premises and burst to the cloud for a true hybrid IT infrastructure. The customer experience will be the same regardless of where the software is running because Weka is cloud-native to start. The mobility of WekaIO reduces the complexity of deploying applications on premise or in the cloud, and results in better resource utilization.


So, what matters in storage? Storage systems deliver the necessary services such as performance, capacity, and availability. Differentiation between systems  derives from features that enable organizations to reduce cost overruns and align storage spend with business value. The innovation driving efficiency and effectiveness stems from the flexibility to leverage the appropriate storage media for the job, enable applications to use services where and when they are needed, and deliver optimal performance per unit of capacity per dollar spent. Next time you evaluate storage systems, remember what matters in storage – that it is the innovative way in which value is delivered at a lower cost that makes a system stand out from the others. That is the differentiator rather than whatever flashy new feature might be promoted with fanfare.

*NAND gets its name from the Logic Gates but refers to the underlying media used in Flash technology.

To learn more about WekaIO storage services, click here.

You may also like:
Worldwide Scale-out File-Based Storage 2019 Vendor Assessment Report
5 Reasons Why IBM Spectrum Scale is Not Suitable for AI Workloads
Comparing Network Attached Storage (NAS) Solutions: Isilon vs. Flashblade vs. Weka

Related Resources

White Papers
White Papers

Top 5 Misconceptions About GPUs for AI


[Webinar] Accelerating AI Training Models

View Now
Case Studies
Case Studies

Preymaker VFX Studio-in-the-Cloud

View Now