Introducing the Intrepid 3000 SSD Series

Identifying the Key Features and Capabilities of OCZ’s Next-Generation, Enterprise-Class, SATA III-Based SSDs

Scott Harlin
Chaitri Sewak

Introduction

IT managers struggle with the best ways to maximize the benefits of flash in their data centers, not only addressing the limitations that hard drive technology imposes on enterprise applications and I/O access, but also in a manner that doesn’t disrupt their proven application usage models. SATA III (Serial Advanced Technology Attachment or Serial ATA) has become the most widely used SSD interface in the industry (following the original SATA and SATA II specifications) and forecast by leading research firms to continue this trend over the next few years. The SATA III standard doubles the maximum data transfer rate over SATA II from 3 to 6 Gigabits per second (6Gbps) enabling faster data transfer speeds between storage units, hard disk drives (HDDs), solid state drives (SSDs), optical drives and protocol host bus adapters (HBAs).

Though the SATA III standard is the prevalent SSD interface, it presents a series of challenges that IT managers must face. One of which includes emerging flash types whereby NAND flash geometries are constantly migrating to smaller lithographic processes creating added pressure on IT departments to be current with the latest trends. Unfortunately, ‘being current’ comes with a price as shrinking flash processes affect NAND reliability, which in turn, reduces SSD life.

IT is also concerned with the wide range of applications driving a variety of mixed workloads, not only from a support perspective, but from a performance perspective as efficient, predictable performance is essential to the success of the enterprise so that connected users are provided with a faster, more responsive computing or storage experience. Without this level of consistency, I/O response latency spikes become evident in the system requiring IT personnel to adjust application workload requirements accordingly increasing maintenance and support costs, as well as overall total cost of ownership (TCO).

Intrepid 3000 SATA SSD
OCZ’s new Intrepid 3000 Series is available in two configurations including the 3600 model supporting MLC read-intensive applications and the 3800 model supporting eMLC write-intensive or mixed workload applications.

The combination of SATA’s strong market position in SSD storage coupled with the IT challenges outlined above has created an opportunity for improved SATA-based storage products and technologies. In understanding these market dynamics, OCZ introduced its next-generation SATA III Intrepid 3000 Series of enterprise-class SSDs (December 2013) delivering the industry’s best sustained performance coupled with predictable, efficient and consistent latency responses.

The series supports the latest 19 nanometer (nm) NAND flash and supports storage capacities up to 800GB representing OCZ’s highest performing and largest capacity enterprise SATA SSDs to date. The architecture is based on OCZ’s Everest 2 platform (that combines a Marvell 88SS9187 controller with proprietary OCZ firmware) and provides advanced flash management and endurance capabilities that extend NAND life and enhance drive reliability.

As mixed workloads are an IT concern, the Intrepid 3000 Series is available in two distinct configurations (as described below) to address both read-intensive and write-intensive applications:

  • The Intrepid 3600 features reliable and cost-effective Multi Level Cell (MLC) NAND media designed for read-intensive applications (such as online archiving, media streaming and web browsing) and guaranteed to complete one drive write per day (DWPD) for 5 years.
  • The Intrepid 3800 features high endurance enterprise MLC (eMLC) NAND media designed for write-intensive or mixed workload applications (such as OnLine Transaction Processing (OLTP), Virtual Desktop Infrastructure (VDI), email servers and analytics) and guaranteed to complete four drive writes per day (DWPD) for 5 years.

Intrepid 3600/3800 models are based on 100GB, 200GB, 400GB and 800GB usable storage capacities, in 2.5-inch industry standard form factors, making the entire portfolio ideally suited for mega data centers and makers of storage appliances where optimum performance, backed by high endurance and reliability, are critical.

The purpose of this white paper is to introduce OCZ’s new Intrepid 3000 Series, including both the Intrepid 3600 and 3800, with concentration on performance, I/O consistency, reliability and endurance, OEM efficiency, and central management. A competitive matrix outlining key capabilities is included.

The Best in Steady-State Performance

Steady state performance (or sustained performance) is performance delivered over time. To measure this requires that the SSD be in a steady state condition having experienced enough program/erase (P/E) cycles so that write operations performance times are stable and can be evaluated in a consistent manner. The SSD erase process is performed one block at a time, at the beginning of each write operation, which adds to the write cycle time.

When an SSD is new, the drive is essentially empty and can perform write operations without having to stop and erase blocks first. This means that a new drive will have faster write times than a used drive and performance metrics during the initial fresh out-of-the-box (FOB) period will be higher and better when compared to a drive that is used or saturated. Therefore, to accurately evaluate SSD write operations performance, pre-conditioning is used to replicate the drive’s steady state and consists of a series of write operation performed in a fixed pattern.

Incredibly fast best describes the new Intrepid 3000 Series. In a steady state condition by which the drive is writing, erasing and re-writing data repeatedly over the full capacity of the drive, the performance for both large block sequential operations, as well as small block random operations, is at the top of its competitive class with specifications that include:

520 MB/s for sequential reads (128K blocks)
470 MB/s for sequential writes (128K blocks)
91,000 IOPS for random reads (4K blocks)
40,000 IOPS for random writes (4K blocks)

{IOPS = input/output operations per second}

Utilizing performance measurements based on Storage Networking Industry Association (SNIA) metrics, and specifications from publicly available published materials, the following chart compares the IOPS performance and bandwidth specifications of the Intrepid 3000 Series versus the Samsung 843T and the Intel S3700, all utilizing 400GB storage capacities in a 2.5-inch form factor.

Intrepid 3000 Performance Specs

Leveraging OCZ’s proprietary firmware, the Intrepid 3000 Series delivers five times faster sustained performance for 4K write operations and two times faster sustained performance for 4K read operations versus OCZ’s previous Deneva 2 SATA III Series of enterprise drives as depicted in Figure 1, and accomplishes these top speeds regardless of whether data is in a compressed or uncompressed format.

Intrepid 3000 vs Deneva 2
Figure 1: The Intrepid 3000 Series delivers faster sustained performance over the previous OCZ SATA III generation

Consistent I/O Access

Latency is the time it takes for a packet of data to get from one designated point to another and includes any delay or waiting incurred that increases the real or perceived response time. In many cases, latency is attributed to mismatches in data speed between the server processor and HDDs as their mechanical heads need to move from location to location limiting their ability to quickly read random data. Within an enterprise network, the data packet may be subject to storage and I/O access delays from intermediate infrastructure devices such as switches and bridges as well.

In the data center, IT managers strive for consistent I/O responses so that predictable and more efficient I/O performance can be achieved in support of the user base. The result of delivering predictable and consistent I/O access across the enterprise not only reduces potential system and storage bottlenecks, but also helps to improve end-user (client) productivity and the overall computing experience. From a TCO perspective, consistent and predictable I/O performance in the data center helps reduce overall power consumption and associated cooling costs, maintenance and support costs, and enables administrative jobs to be scheduled and performed with confidence.

A good example of the effects of consistent and predictable I/O performance is evident when comparing the new Intrepid 3000 Series platform to OCZ’s enterprise-class Deneva 2 architecture. In an identical benchmark test performing a series of small 4K block write operations, the Intrepid 3000 Series consistently delivered I/O response times of 0.028ms over a sustained period. Under the same test scenario, the Deneva 2 Series delivered I/O responses that ranged from 0.32ms to 0.37ms.

The Result: the Intrepid 3000 Series provided a 12x or 1200% improvement in I/O response time as depicted in Figure 2 below, delivering consistent and predictable latency over a sustained time period.

Average Latency (4K Write)
Figure 2 represents average latency results that compare the new Intrepid 3000 Series to OCZ’s previous Deneva 2 Series when 4K block write operations are performed over a sustained period of time.

Enterprise-Class Endurance and Reliability

With a host of write-intensive and mission-critical applications driving the data center, enterprise SSDs are expected to always be available for moving data to new locations based on need (or tier structure), for providing a performance boost to key applications, or simply for enabling ‘real-time’ data access. The life expectancy of an SSD is completely dependent on the endurance and reliability tools, techniques, algorithms and/or technologies implemented within its SSD controller.

Enterprise SSD endurance is typically described as the built-in capabilities that enable the device to handle a constant workload without stress, malfunction or failure for a long period of operation with emphasis on those capabilities that extend NAND flash life and improve drive durability. Data reliability includes those attributes built within an SSD that enable consistent ‘to-spec’ performance and provide advanced data recovery, data integrity and power loss protection capabilities.

The Intrepid 3000 Series features OCZ-engineered proprietary firmware that works on top of the proven Marvel 88SS9187 controller to deliver an advanced suite of endurance and reliability tools designed to extend NAND flash memory life while providing the enterprise-class endurance, reliability, data integrity and power loss protection capabilities required by today’s data center managers. The following includes some of these key capabilities:

Intrepid Key Features

Advanced Multi-level BCH Error Correction Code

The most widely used ECC algorithm for enterprise SSDs are the Bose-Chaudhuri-Hocquenghem (BCH) codes whose algorithms generate extra parity bits to enable the controller’s ECC engine to perform required detections and corrections in real-time. The more bits of ECC used results in higher levels of error correction, which in turn, improves data reliability. Despite whatever wear and tear the drive has experienced, ECC attempts to correct and deliver reliable data.

The Intrepid 3000 Series provides advanced levels of ECC correcting errors up to 85 bits per 2Kb of data while significantly reducing the unrecoverable bit error rate (UBER). This advanced ECC is more robust and reliable than traditional ECC algorithms and enable either one complete drive write per day (DWPD) for 5 years (Intrepid 3600 Series) or four complete DWPD for 5 years (Intrepid 3800 Series).

End-to-End Data Path Protection

To extend error detection and correction, the Intrepid 3000 Series performs data integrity checks at every juncture where data is transmitted, received, processed and stored, to ensure that corrupted data will be detected immediately and not propagated within the enterprise. The Intrepid 3000 Series SSD controller monitors each device in the data path (from the host to DRAM to the controller and onto SSD flash), detecting and reporting errors before the data is written to the Intrepid 3000 Series device and utilizes SATA cyclic redundancy checking (CRC) as part of the process. CRC compares sending data to receiving data making sure that data has been received successfully.

In-flight Data Protection

The loss of in-flight mission-critical data and metadata can be highly detrimental to any business. To prevent data loss from sudden power drops, outages or inadvertent disengagements and to ensure superior data integrity, in-flight data protection must be implemented at the storage level.

The Intrepid 3000 Series includes advanced in-flight data protection, so if a sudden power loss occurs and the primary power source drops below a predefined threshold, the Intrepid 3000 Series SSD controller will no longer allow the drive to accept new commands from the host until the in-flight write operation completes. Once that occurs, power loss backup circuitry is activated to enable the in-flight data to be safely transferred and stored in flash.

Data Redundancy via SSD RAID

Redundancy is a process that duplicates or replicates all data stored in a flash-based SSD so if a drive fails, backup exists and the storage operation will not be adversely impacted so that users will be confident that their data is protected under most failures or duress. The ability to provide RAID-like protection within the Intrepid 3000 Series is achieved by enabling the data blocks to be distributed amongst multiple NAND flash cells, generating parity information as well. As a result, Intrepid 3000 Series drives achieve faster rebuilds utilizing a smaller stripe size as longer rebuilds can cause system performance degradation over time.

Creating this level of redundancy through block-level striping and parity distribution across multiple flash cells safeguards against NAND flash failures and uncorrectable data errors beyond what ECC covers, providing yet another mechanism for extending the Intrepid 3000 Series NAND flash life, saving time and money while providing user confidence that their data is recoverable and accurate.

Advanced Flash Management

The Intrepid 3000 Series implements a host of advanced flash management capabilities (i.e. wear-levelling, garbage collection, and support for TRIM), that not only manage flash resources but perform operations that free up flash cells for high-performance storage. For example:

Wear-levelling extends Intrepid 3000 Series drive life and improves overall drive endurance and is an intelligent form of block management that organizes data in a manner so the program cycles are evenly distributed amongst all flash cells.

By doing so prevents flash from being overused, which in turn, slows cell wear. Garbage collection is another process featured within the Intrepid 3000 Series by which the SSD controller procedurally determines which flash cells have unneeded data and either consolidates or erases those blocks to reclaim usable capacities for use in normal operations. By erasing unneeded data in flash reduces cell use that equates to improved drive life.

When a file is permanently deleted within an Intrepid 3000 Series SSD, the operating system sends a TRIM command along with the Logical Block Addresses (LBAs) that no longer contain valid data to the SSD controller. This informs the controller that the LBAs in use can be erased and re-used, enabling each Intrepid 3000 Series SSD with more capacity to store data and higher performance.

SSD Health Monitoring

The Intrepid 3000 Series can monitor cell usage to predict flash deterioration and life expectancy through SMART Technology and StoragePro Central Management. This level of proactive health monitoring enables the SSD controller to analyse and make adjustments dynamically to whatever volatilities the Intrepid 3000 Series SSD is experiencing as the flash ages, and predicts how read and write operations will affect drive performance in the future.

Implemented at the system level is a watchdog technology that manages and evaluates access between each Intrepid 3000 Series SSD and the host’s basic input/output system (BIOS). Based on past and present use-analysis it receives from the Intrepid 3000 Series SSD controller, this industry-standard, self-monitoring analytical and reporting (SMART) technology initiates warning messages to the controller about potential drive failures enabling IT administrators to respond in a proactive manner.

To take SSD monitoring and management to new levels, the Intrepid 3000 Series is supported by OCZ’s StoragePro XL 1.1 application that enables IT managers to centrally monitor and administer connected Intrepid 3000 Series flash resources from a web-based management interface. This networkaccessible management system securely connects to multiple hosts (running Linux or Windows operating systems) providing a cross-platform view of its flash resources connected to servers, storage arrays or appliances.

With StoragePro, IT managers are afforded specific drive details on performance, reliability and operation. Along with the monitoring functionality, a user configurable alerting systems is provided that enables identification of any potential system and/or storage issues in advance enabling IT corrective actions to be initiated at an early stage. The user-friendly StoragePro XL GUI provides:

  • A structured group-based view of host and Intrepid 3000 Series SSD activity throughout the data center
  • Critical alert displays and warnings from hosts and connected Intrepid 3000 Series SSDs
  • Simpler and easier Intrepid 3000 Series SSD installation, management and maintenance
  • Fast and easy routine Intrepid 3000 Series SSD maintenance runs, host system checks and administrative tasks from firmware updates to printing detailed reports

StoragePro remote host and SSD management provides the system information and SSD health that IT professionals need to centrally perform mission-critical actions and maximize data center ROI from their enterprise flash resources.

OEM-Optimized

As previously mentioned in this paper, the Intrepid 3000 Series is ideally suited for storage manufacturers where optimum performance backed by high endurance and reliability are critical. Supported by SMART attributes that provide advanced product health monitoring and StoragePro XL central management, the Intrepid 3000 Series is optimized for OEM usage enabling better and more efficient management and maintenance so that OEMs can confidently implement these SSDs into OEM products (under OEM brand names).

Intrepid OEM-Optimized

In addition to these management and monitoring capabilities, the Intrepid 3000 Series includes enhanced logging capabilities for proactive monitoring. This capability provides a deeper dive down of SSD failures that have occurred with more information as to the cause of the problem. The failures are data logged on Intrepid 3000 drives, analysed and the results can be saved, shared or outputted. This level of data logging is crucial to OEMs as this process collects information faster and more accurately than a manual process.

The OEM qualification process can be a daunting and lengthy activity requiring a set procedure and elaborate testing to assure that product/system specifications are met as required by the manufacturer as the process confirms performance, quality, compliance, endurance, reliability, cost and the on-time delivery of components, to name a few. Typically, a new product needs to be qualified and approved before it can be implemented within an OEM design. A new version of a qualified product typically must be re-qualified by the OEM as well. To ensure a consistent solution, the Intrepid 3000 Series supports locked firmware as no changes are allowed streamlining the OEM qualification process.

Summary

The new Intrepid 3000 Series leverages OCZ in-house firmware with an impressive enterprise feature-set to enable customers with unprecedented performance, data management, endurance and reliability, and cost efficiency resulting in an optimal flash-based storage environment. It delivers exceptional SSD responsiveness in even the most demanding and compute-intensive applications achieving unsurpassed performance under any workload regardless of data type or I/O pattern.

Intrepid 3000 Series SSDs overcome today’s data integrity and performance inconsistency challenges and enable users to flash-optimize their businesses and/or applications seamlessly, and with complete confidence. With an advanced enterprise-grade feature-set, the Intrepid 3000 Series delivers leading performance, endurance and reliability ideal for mega data centers and storage appliance OEMs.

In comparison to other SATA III-based enterprise SSDs currently available, the Intrepid 3000 Series is at the top of its class in sustained performance as outlined from the competitive matrix below. Utilizing performance measurements based on Storage Networking Industry Association (SNIA) metrics, and specifications from publicly available published materials, the competitive matrix compares sequential read and write throughput based on larger 128K block sizes, random read and write IOPS performance based on smaller 4K block sizes, design writes per day (DWPD) endurance guaranteed, wattage of power consumed and the NAND flash lithography used.

Intrepid 3000 Comparison Chart

The Intrepid 3000 Series provides industry-fast, efficient and predictable performance that results in a more responsive environment for IT applications, and with configurations designed for both read-intensive and write-intensive/ mixed workload environments, it supports the complete spectrum of enterprise applications including online archiving, media streaming, web browsing, OLTP, VDI, email and analytics.new version of a qualified product typically must be re-qualified by the OEM as well. To ensure a consistent solution, the Intrepid 3000 Series supports locked firmware as no changes are allowed streamlining the OEM qualification process.

DOWNLOAD PRINTABLE VERSION (PDF)