OIF 448Gbps Signaling for AI Workshop

Empirical file and disk I/O operation performance metrics can be invaluable with regard to substantiating theories and assessing claims about disk I/O performance. This is especially so when these metrics reflect the actual file and disk I/O operation activity performed by individual applications and workloads during normal usage.

SSD performance depends on the type of workload applied to the storage, i.e. performance depends on the application workload. In this survey, we examine various real world user workloads for both client and enterprise applications. See how different workloads affect SSD performance. See the impact of data reduction, virtualization, encryption and other in-line and SSD based optimizations.

Learning Objectives

What are realworld workloads
What affects workloads
How do different workloads affect perfromance

SSD performance depends on the workload that it sees. IO streams are constantly changed as they traverse the IO stack such that IO streams generated in user (application) space differ from what the storage sees. Here, we look at a variety of real world workload captures and how these workloads affect SSD performance.

Learning Objectives

Understanding what real world workloads are
Seeing how real world workloads affect storage performance
Comparing workloads across different SSDs

While originally designed for disk drives, the read/write I/O model has provided a common storage abstraction for decades, regardless of the type of storage medium. Devices are becoming increasingly complex, however, and the constraints of the old model have compelled the standards bodies to develop specialized interfaces such as the Object Storage Device and the Zoned Block Commands protocols to effectively manage the storage. While these protocols have their place for certain workloads, there are thousands of filesystems and applications that depend heavily on the old model.

Optimized storage performance through available configuration parameters can be challenging. Too many parameters can be tuned and with tuning one-factor-at-a-time (OFAT) by developers, operators and engineers, the overall cost of test and time is inefficient and ineffective. This presentation introduces design of experiment (DOE), a method systematically designing the experiment and mathematically calculating the tested parameter’s effects toward storage performance. A case study using 2 types of storage technologies, all HDD and all SSD as Ceph storage will be discussed.

Security compliance of unstructured data has become an ubiquitous business requirement and even so more with the upcoming GDPR regulation. Object Store is growing continually in deployments hosting oceans of unstructured data. Understanding which object data falls under compliance governed data category becomes vital so that the required security compliance enforcement from the storage side can be acted on it.

NetApp®WAFL®is a transactional file system that uses the copy-on-write mechanism to support fast write performance and efficient snapshot creation. However, copy-on-write increases the demand on the file system to find free blocks quickly; failure to do so may impede allocations for incoming writes. Efficiency is also important, because the task may consume CPU and other resources.

As the Storage Performance Development Kit (SPDK) continues to grow in popularity in both open source and proprietary projects in the storage industry, the SPDK Community continues to evolve and our focus areas continue to adapt to community needs. As many are already be aware, SPDK is an open source set of tools and libraries for writing high performance, scalable, user-mode storage applications. It achieves high performance by moving all of the necessary drivers into user space and operating in a polled mode instead of relying on interrupts.

Your typical SSD is very badly behaved with regard to latency. Background tasks interfere with I/O requests. As drives increase in size beyond a typical working set, multiple users interfere with each other (noisy neighbor). Scale out systems such as those assembled by Hyperscalers, are not tolerant of today’s long tail of SSD latency. In order to address these requirements, the storage industry through the NVM Express standard is adding some new features. This talk will explain those features and how they address these issues.