Nantero NRAM™ is a new class of memory with the potential to add non-volatility to existing RAM applications. It can be arranged in a crosspoint structure for large memories or a 1T-nR arrangement for smaller faster arrays, in standalone devices or as embedded RAM. NRAM uses carbon nanotubes in a dielectric-free structure to achieve unlimited write endurance. While there are obvious advantages to this class of device, including replacing DRAM in storage devices, there are a number of less obvious changes to how designers approach the data storage hierarchy.

Computer architecture is embarked on several new paths including quantum computing, adiabatic computing, approximate computing and several approaches to bringing processing closer to the memory. These new architectures will allow important developments in Deep Learning, Artificial Intelligence, and machine learning which will facilitate the development of advanced robotics, industrial IoT, autonomous vehicles, and other applications. A common link between all of these is their need for storage and memory.

From cloud systems to local machines, SSD and Flash to disk drives, components to systems, and hardware to software defined storage, innovations in storage system technology are constantly being developed by companies both large and small. One key differentiator that innovative and cutting-edge storage companies can leverage to keep ahead of their competition is intellectual property protection. This presentation will cover patent protection for various storage technologies, with an in-depth look at both hardware and software.

In a multi-tier virtualized storage environment, there is continuous movement of data from higher tiers (flash) to lower tiers. If this is slow, higher tiers may run out of space due to high workload ingest rates. Especially in multi-tenant cloud environment, high-ingest behaviour of certain workloads may undesirably affect other high priority workloads. In this presentation, we will detail how using storage QoS, total workloads' ingest rate can be made proportional to the residual space on higher tier and the remaining IOPS capacity can be reserved for flush.

For cloud-based storage and compute systems optimal performance is possible when one understands characteristics of workload and dynamically take action based on the needs. To detect and classify mix of workloads for large number IO-streams in real-time is thus essentially a pattern recognition problem.

In modern data center with thousands of servers, thousands of switches and storage devices, and millions of cables, failures could arise anywhere in compute, network or storage layer. The infrastructures provides multiple sources of huge volumes of data - time series data of events, alarms, statistics, IPC, system-wide data structures, traces and logs. Interestingly, data is gathered in different formats and at different rates by different subsystems.

A storage ecosystem is comprised of several servers which could be heterogeneous with components from multiple vendors. The administrator for this ecosystem should be able to obtain information about each intelligent component connected, and also manage them, without worrying about intricacies of communication with that component. Standardization of the way the component and its data are represented, helps achieve interoperability which the administrator needs. SMI-S, Redfish and Swordfish are standard specifications which model hardware components of a storage ecosystem.

While the concept of a log is certainly not a new concept in computer science, until recently logs have been used as a part of an application, typically for crash recovery purposes or sometimes for auditing/debugging purposes. More recently, logs have been emerging as a first class storage concept in and of themselves, being used in distributed environments as a mechanism for communication, as a mechanism of persistence and recovery for services, and as an enabler for query optimized data structures in complex systems.

Assessing the impact to storage capacity, throughput, and storage companies

Ozone brings in a new storage paradigm in Hadoop called object storage. It will co-exist with HDFS to provide file store and object store functionality in the same Hadoop cluster. Ozone will also solve the scalability and small file problem of HDFS, where users can now store trillions files in Ozone and access them as if they are on HDFS. Ozone plugs into existing Hadoop deployments seamlessly and programs like Hive and Spark work without any modifications. This talk looks at the architecture, reliability, and performance of Ozone.