Kraftway presents a new product in the hardware-oriented cluster storage segment: trusted storage systems PROGRESS, which are based on controllers with integrated information security tools and Russian software products disk array management. Distinctive features trusted storage platform Kraftway:

  • built-in information security functions integrated into the controllers (circuitry of the board, BIOS and firmware code of the Navy);
  • software(Software) for managing the storage system, entered in the register of the Ministry of Communications of the Russian Federation. The software is produced by Radix, NPO Baum and Aerodisk.

Trusted storage is designed for customers who have specific security requirements for their IT systems. DWH PROGRESS contains models with the number of controllers 1,2,4 and 8 (*) operating in Active-Active mode and providing high fault tolerance of the storage system. Expansion disk shelves of various standard sizes are connected to the system controllers, providing a storage capacity of up to several tens of PB. The maximum capacity of 2-controller storage is 16 PB. Storage host interfaces: iSCSI 1 to 100 Gb/s, FC 2 to 32 Gb/s, Infiniband up to 100 Gb/s (*).

(*) Specifications may vary for each software vendor.



  • E. Access control


    Attached file

  • Trusted storage systems Kraftway PROGRESS

    Ask a Question

  • Trusted storage systems Kraftway PROGRESS

  • Trusted storage systems Kraftway PROGRESS

    Features of Kraftway PROGRESS storage management software

  • Trusted storage systems Kraftway PROGRESS

    Main scenarios for the use of VAZ

    A. Ensuring that only authorized personnel have access to the storage controller

    For loading operating system controller requires two-factor authentication. After the power is turned on, the VZZ stops the download process. An authorized user must have an identification device (smart card, USB key) and a password to proceed with the download.
    VZZ has the ability to differentiate the rights to manage security settings depending on the user's role. Regular user, for example, may not have permission to enter and change UEFI BIOS settings.

    B. Controlling the integrity of the hardware configuration

    After power is applied, the VZZ performs self-testing, checksum calculation and comparison with the reference ones. If successful, the integrity of the equipment is monitored by comparing checksums and signaling when changes are detected. If the integrity is violated, only a user with Administrator rights will be able to manage the VZ.

    B. Controlling the integrity of the file system

    The VZ administrator can enable checking the integrity of critical files for changes. In this case, when the product is turned on before the OS is loaded, the checksums of the files added to the control list are calculated. If the integrity is violated, only a user with Administrator rights will be able to manage the integration VIS

    D. Antivirus scanning before the operating system starts

    Searching for malware at the UEFI stage before the operating system boots allows you to neutralize threats that are extremely difficult to detect after the OS starts, the so-called "rootkits" and "bootkits". They can modify boot sectors system, as well as to hide traces of the presence of an attacker or malware in the system. The search is carried out by a specialized module "Kaspersky Anti-Virus for UEFI". In case of discovery malicious code the scanner pauses OS loading and identifies the infected object.

    E. Access control to hardware resources using a "thin hypervisor". The hypervisor is part of UEFI and is a software tool for restricting access to hardware resources of a computing device.
    The hypervisor functions in the virtualization mode of all those physically present on the motherboard input/output devices, as well as input/output ports and direct memory access channels. The hypervisor provides access control to external media, including a ban on their use, as well as centralized accounting of connected removable media.

    Storage management software features

    The attached file contains a description and features of disk array management software from each of the manufacturers: Radix, NPO Baum and Aerodisk.

    • Dell EMC Storage SC Series are automated, modern infrastructure solutions built with hybrid storage and high-end Flash arrays.
    • The Dell EMC Equallogic PS Series are the ideal devices for the corporate information environment, enabling the efficient implementation of day-to-day information tasks.
    • The Dell POWERVAULT MD Series are scalable, low cost systems that support the consolidation of large amounts of data and simplify data management.
    • EMC VNXE Series are unified storage solutions for small business information needs.

    Entry level storage

    Dell EMC entry-level storage systems provide high-performance platforms for small businesses as well as large companies that are distinguished by their extensive branch infrastructure. This class of hardware is highly scalable from 6 to 150 drives for a maximum storage capacity of 450TB. Dell EMC storage systems are ideal for enterprises with advanced infrastructure of physical server systems, as well as for those who practice the use of virtualized server systems. The practical use of Dell EMC storage will allow you to consolidate large amounts of information, as well as improve the efficiency of their processing. Using these devices, it will be possible to deploy multifunctional storage systems based on IP networks that support file and block access protocols, respectively, NAS and iSCSI.

    Midrange storage

    Dell EMC Midrange Storage is a feature-rich platform that enables consolidation of block storage, file server systems, and direct-attached storage. The use of this equipment will enable companies to dynamically grow file systems and block resources with parallel support for several protocols - NFS and CIFS. In addition, storage stores can provide access to information using protocols such as Fiber Channel, iSCSI, and FCoE. This will help support block applications that require high bandwidth and low latency.

    Until a few years ago, expensive dedicated hard drive storage systems were focused mainly on mission-critical enterprise applications or some specific tasks. Today, thanks to the rapid development of the NAS (Network Attached Storage) concept and a number of other solutions based on ATA hard drives, such entry-level systems are becoming an object of attention for medium-sized enterprises, which is especially important for the price-sensitive Ukrainian market.
    There are enough serious reasons for the interest of buyers in storage systems
    data - for example, the need to consolidate information arrays, the need
    solve data redundancy problems in large networks, etc. On the other hand,
    they appear during the operation of high- and medium-level systems.

    According to a survey of several Kyiv companies, today the following motives for the acquisition of drives are most often encountered (in descending order).

    1. Additional disks impossible or uneconomical to install in a server (usually either due to lack of space in the chassis, or high price to original disks, or a non-mass OS and platform configuration - for example, Silicon Graphics or Compaq Alpha Server, Mac, etc.).

    2. You need to build a failover cluster with a shared disk array. In such a situation, you can sometimes do without a storage system, for example, by using PCI-SCSI RAID controllers with support for cluster systems, but this configuration is less functional and, moreover, does not allow you to enable data write caching in the controllers. When working with databases, the performance of solutions with an independent storage device sometimes outperforms systems based on PCI-SCSI RAID controllers by an order of magnitude.

    3. It is impossible to get a quality storage solution within a standard server. An external system in this case allows you to implement RAIS (Redundant Array of Independent Servers - a fault-tolerant array of independent servers). It stores everything, including system data, accessed by the servers that process it. At the same time, a spare server is provided that replaces the failed one. This approach is somewhat similar to clustering, but it does not use specialized software, and applications do not automatically migrate.

    Common classification of data storage systems
    based on the principle of organizing access to them.


    SAS (Server Attached Storage)— drive connected to the server.
    The term "direct-attached drive" is sometimes used—
    DAS (Direct Attached Storage).

    The main advantage of a drive connected to a server compared to
    with other options low price and high speed.


    NAS (Network Attached Storage)— a drive connected to the network.

    The main advantage of this solution is the speed of deployment and thoughtful
    organization of access to files.

    SAN (Storage Area Network) - storage network.
    The most expensive solution, at the same time providing many advantages
    - independence of the SAN topology from storage systems and servers, convenient
    centralized management, no conflict with LAN/WAN traffic, convenient
    backing up and restoring data without loading the local network and servers,
    high speed, scalability, flexibility, availability and fault tolerance.

    Storage systems, or stand-alone drives

    Definitely a drive or storage system
    a lot of data can be various devices. But as soon as we have a speech
    will talk about disk systems that provide storage of information and access to it,
    we will mean by the term "accumulator" exactly them. All in all
    case, they consist of hard drives, an I / O controller, and an integrated
    systems. Drives are usually hot-swappable, i.e. they are
    can be connected and disconnected "on the fly", without turning off the drive. it
    makes it possible to replace a failed hard drive without any trouble
    for the user. The drive's primary and backup power supplies have increased
    reliability and are also hot swappable. Yes, and I / O controllers
    sometimes two are used. Diagram of a typical disk storage system with one
    controller can be seen in Fig. one.

    The controller of the disk storage system is its center. He is responsible for input / output of data within the system and to external channels, as well as for organizing the storage and access to information. To communicate with the outside world, drive controllers usually use SCSI, Fiber Channel, or Ethernet interfaces.

    Depending on the purpose of the system, the controllers can implement different operation logic and use different data exchange protocols. They provide user systems with data at the block level, like hard drives, or file services using the NFS, CIFS protocols, as well as Network File System, Common Internet File System like file servers (see sidebar "File protocols in NAS - CIFS, NFS, DAFS"). Such a controller typically supports standard RAID levels to increase system performance and provide fault tolerance.

    File protocols in NAS
    - CIFS, NFS, DAFS     CIFS (Common Internet File System) is a standard protocol that
    provides access to files and services on remote computers (including
    including on the Internet). The protocol uses a client-server interaction model.

    The client makes a request to the server to access files or send a message
    program on the server. The server fulfills the client's request
    and returns the result of its work. CIFS is an open standard developed
    based on Microsoft's SMB protocol (Server Message Block Protocol), which
    traditionally used in local networks with Windows OS to access files
    and print. Unlike the latter, CIFS is focused on application in that
    number and in distributed networks - for example, it takes into account the possibility of
    big timeouts.

    CIFS uses TCP/IP to transport data. It provides the functionality
    similar to FTP ( File Transfer Protocol), but provides clients with an improved
    (similar to direct) file control. It also allows sharing
    to files between clients by applying blocking and automatic recovery
    communication with the server in the event of a network failure.

    NFS (Network File System) is an IETF standard that includes distributed
    file system and network protocol. NFS was developed by Sun and
    originally used only on Unix systems. Later implementation of the client
    and server parts have spread to other systems.

    NFS, like CIFS, is based on a client-server interaction model. It provides
    access to files on a remote computer (server) for writing and reading
    as if they were on the user's computer. In earlier versions
    NFS used the UDP protocol to transport data, in modern
    — TCP/IP. Sun developed a protocol for NFS operation on the Internet.
    WebNFS, which uses extensions to the NFS functionality to correctly
    work on the World Wide Web.

    DAFS (Direct Access File System) is a standard file access protocol,
    which is based on NFSv4. It allows application tasks to transfer
    data bypassing the operating system and its buffer space directly
    to transport resources, preserving the semantics file systems. DAFS uses
    Benefits the latest technologies data transfer according to the "memory-memory" scheme.
    It provides high file I / O speeds, minimal download
    CPU and entire system thanks to a significant reduction in the number of operations
    and interrupts, which are usually needed in the processing of network protocols.
    Particularly effective is the use of hardware support
    VI (Virtual Interface).

    DAFS was designed for continuously running databases and a variety of
    Internet applications in a cluster and server environment. It provides
    the lowest access delays to shared file resources and information, as well as
    supports intelligent recovery mechanisms
    system and data, which makes it very attractive to use
    in high-end NAS drives.


    Why ATA?

    Today, the difference in cost per unit volume of large ATA and SCSI drives is more than
    than six times, and this ratio is quite justified. Expensive interface discs
    SCSI is intended primarily for corporate information systems and usually
    have more high performance speed when processing large quantities
    requests. They use more reliable components, they are better tested, yes
    and the responsibility of the manufacturer for these devices is much higher.

    But if the cost of data is not so high or only an intermediate device is needed
    when reserving them, why pay six times more? Considering that the exit
    building one of the disks in the array is not critical, it is perfectly acceptable to use
    drive with ATA disks. Of course, there are a number of contraindications for the use
    ATA drives in large storage systems, but there are also a number of applications
    for which they are perfect.

    IDE devices are most widely used in entry-level NAS systems. When using two or four disks organized in a RAID 1 or 0 + 1 array, the probability of failure of the entire system is acceptably small, and the performance is quite enough "with a head" - file servers entry-level drives don't do too many disk operations per second, and data flows are limited to external Fast Ethernet or Gigabit Ethernet interfaces.

    Where block access to data is required at the minimum cost of the solution and
    the number of operations per unit of time is not a critical parameter, use
    systems with an external parallel SCSI or Fiber Channel interface and ATA drives
    inside (Fig. 2).

    Leading manufacturers today offer ATA drives that are close in all characteristics,
    including MTBF to industrial SCSI drives. Together with
    the more their cost becomes comparable and, accordingly, the use of ATA-disks
    provides only a small gain in the price of drives.

    For entry-level servers and workstations that store enough
    important data, the use of cheap PCI ATA controllers, as practice shows,
    does not always give the desired result due to their relative primitiveness and small
    functionality. The use of expensive external drives is not always justified.
    In this case, you can use the ATA-to-ATA device, which is a reduced
    a copy of an external disk storage system and is designed for only two disks
    with ATA interface. However, it has a fairly high-quality built-in controller.
    and supports "hot-swap" drives (Figure 3).


    Serial ATA - a new breath of the ATA interface

    With the advent of the Serial ATA interface of data storage systems on ATA disks
    should become more. Almost all drive manufacturers talk about this.
    entry level. Today, their new models are already equipped with a new interface. How
    Is the Serial ATA interface interesting for manufacturers of data storage systems?

    It supports the Native Command Queuing instruction set (command pipelines) - the controller analyzes I / O requests and optimizes the order in which they are executed. True, unlike the traditional Native Command Queuing in SCSI drives, which provided a queue of up to 256 commands, Serial ATA will support a queue of up to 32 commands. "Hot-swapping" of Serial ATA drives, which used to require certain technical tricks, is now written directly into the standard, which will allow creating high-level corporate solutions. The new design is also important: the cable in the new interface has become round, and its connector is small and narrow, which facilitates the design and assembly of systems.

    In new versions, the speed of Serial ATA will increase, and there is no doubt that the share of ATA solutions in entry-level storage systems will increase precisely due to new drives with this interface, while the development of Parallel ATA will slow down, which has been observed recently .

    RAID (Redundant Array of
    independent disks)     Entry-level drives typically use RAID 0 levels,
    1, 5 and combinations of them.

    RAID 0


    Disk array without striped fault tolerance (Striped Disk
    Array without Fault Tolerance). In this case, the data is divided into blocks,
    written in parallel to different disks that are jointly participating in
    every I/O operation.

    The advantage of this approach is to provide high performance for
    applications that require a large amount of I / O data, ease of implementation
    and low cost per unit volume. The main disadvantage is not fault-tolerant
    solution: the failure of any one disk entails the loss of all
    array data.

    RAID 1


    Disk array with duplication. "Mirror" (mirroring)
    - a traditional way to increase the reliability of a small disk array
    volume. In the simplest version, two disks are used, on which
    the same information. In case of failure of one of them, a double remains, which
    continues to work as before.

    Advantages - ease of implementation and restoration of the data array, as well as
    high enough speed for high-intensity applications
    requests. Disadvantages - low data transfer rate at double the cost
    per unit volume, since there is a 100% redundancy. With more
    number of disks, you can use RAID 0+1 or RAID 10 instead of RAID 1,
    combinations of RAID 0 and RAID 1 to achieve the best performance
    system speed and reliability.

    RAID 5



    failsafe
    an array of independent data disks with distributed parity (Independent Data disks
    with distributed parity blocks).

    The data is split at the block level. Each block of data is written to
    specific disk and can be read separately. For data blocks it is counted
    parity and is distributed cyclically across all disks in the array. If operations
    records to be scheduled properly, it becomes possible to parallel
    processing up to N/2 blocks, where N is the number of disks in the group. It increases
    performance, and to obtain a fault-tolerant array, it uses
    just one redundant drive.

    RAID 5 provides high speed writing and reading data, which increases
    performance under high intensity read/write requests and
    This reduces the overhead for implementing redundancy. However,
    its organization is quite complex, and data recovery can be
    a certain problem.


    Serial Attached SCSI

    The SCSI interface has high speed and reliability, but such solutions
    quite expensive. SAS (Serial Attached SCSI) is an interesting evolution of SCSI
    and, in all likelihood, will also be used in low-cost entry-level systems.
    and middle level.

    Today, many storage manufacturers use the Ultra 320 SCSI interface when designing relatively simple drives. This is the generation of the parallel SCSI interface on this moment last in the line. Drives with the previously announced Ultra 640 SCSI interface, most likely, will not be mass-produced or will disappear altogether from the scene. At a recent meeting with partners, Seagate, a leader in enterprise-level hard drives, announced that new drive models for high-end systems will be equipped with a Fiber Channel interface, and for smaller corporate systems- Serial SCSI. At the same time, the usual parallel Ultra 320 SCSI will not disappear immediately. Its final replacement is expected no earlier than in five years.

    Serial SCSI combines some of the features of Serial ATA and Fiber Channel. It was developed from the serial ATA specifications and improved. Thus, the signal level has increased, which allows you to accordingly increase the maximum length of the four-core cable to 10 m. This two-channel point-to-point interface operates in full duplex mode, can serve up to 4096 disk devices in the domain and supports standard set SCSI commands.

    At the same time, despite all its advantages, Serial Attached SCSI is unlikely to replace the conventional parallel interface in the near future. In the world of enterprise solutions, development is carried out very carefully and, of course, longer than for desktop systems. Yes, and old technologies do not go away very quickly, since they life cycle is several years. The first devices SAS interface should be on the market in 2004. Naturally, at first it will be mainly disks and PCI controllers, but data storage systems will appear quite quickly. Comparative characteristics interfaces are listed in the table "Comparison of modern disk interfaces".


    SAN - Storage Area Networks

    SAN (see sidebar "Classification of data storage systems - DAS / SAS,
    NAS, SAN") based on Fiber Channel allow you to solve almost any task
    data storage and access. But there are several drawbacks that negatively affect
    on the spread of these technologies, first of all - the high cost of solutions
    and the complexity of building geographically distributed systems.

    There is a fierce debate surrounding the use of IP protocol as a transport for SCSI commands and data in SANs, but everyone understands that IP Storage solutions will definitely find their niche in the field of storage systems, and this will not be long in coming.

    As part of the improvement of network storage technologies, the Internet Engineering Task Force (IETF) organized a working group and an IP Storage (IPS) forum in the following areas:

    FCIP - Fiber Channel over TCP / IP, a tunnel protocol based on TCP / IP and designed to connect geographically distant FC SANs without any impact on FC and IP protocols;

    iFCP - Internet Fiber Channel Protocol, a protocol for connecting FC systems or storage networks based on TCP / IP, using an IP infrastructure in conjunction with or instead of FC switching and routing elements;

    iSNS - Internet Storage Name Service, storage name support protocol;

    iSCSI stands for Internet Small Computer Systems Interface, a TCP/IP-based protocol designed to communicate with and manage storage systems, servers, and clients.

    The most rapidly developing and interesting of these areas is iSCSI, which became the official standard on February 11, 2003. Its development should significantly affect the spread of SANs in small and medium-sized businesses, due to the fact that storage networks will become much cheaper. As for the use of iSCSI on the Internet, today FCIP has already taken root well here, and competition with it will be quite fierce, but due to an integrated approach, it should work in favor of iSCSI.

    Thanks to IP Storage technologies, including iSCSI, storage networks have new opportunities for building geographically distributed storage systems. In addition, new storage systems that natively use iSCSI will provide many other benefits, such as QoS support, high levels of security, and the ability to use Ethernet specialists to maintain networks.

    One of the very interesting features of iSCSI is that you can use more than just media, switches and routers to transfer data on an iSCSI drive. existing networks LAN/WAN, but also conventional network adapters Fast Ethernet or Gigabit Ethernet on the client side. But in fact, due to some difficulties, it is better to use specialized equipment, which will lead to the fact that the cost of solutions will begin to catch up with traditional Fiber Channel SANs.

    The rapid development of storage networks became the basis for the formation of the concept of World
    Wide Storage Area Network. WWSAN provides for the creation of an infrastructure that
    will provide high-speed access and storage of data distributed around the world.

    Comparison of modern disk interfaces

    Options
    		 Serial
    ATA         		SCSI
    		 SAS
    		 FC
    Number of supported devices 16 16 4096 2 24
    Maximum cable length, m 1 12 10 Copper: 30
    Optics: 10,000*
    Supported topologies dot-dot Tire dot-dot Ring**
    dot-dot
    Speed, MBps 150, 300 320 150, 300 100, 200, 400
    full duplex — — + +
    Interfaces ATA, SCSI SCSI ATA, SCSI Independent***
    Dual port device support — — + +

    * The standard regulates the distance
    up to 10 km for single-mode fiber, there are implementations for data transmission
    over a distance of more than 100 km.
    ** As part of the internal topology of the ring, hubs and FC switches operate,
    there are also implementations of switches that provide a point-to-point connection
    any devices connected to them.
    *** There are device implementations for interfaces and protocols SCSI, FICON,
    ESCON, TCP/IP, HIPPI, VI.

    Infortrend ESDS 1000 Series

    Review Infortrend ESDS 1000

    The EonStor DS 1000 storage systems provide an excellent price/performance ratio. For users of small medium...

    Infortrend ESDS 1000 Series Storage

    The Infortrend ESDS 1000 Series is affordable storage with built-in iSCSI and optional FC/SAS interfaces for enhanced performance and scalability.

    Review Infortrend ESDS 1000

    The EonStor DS 1000 storage systems provide an excellent price/performance ratio. For small medium business (SMB) users, an entry-level solution is provided. Models are available for a different number of HDD drives in various form factors: 12-slot 2U, 16-slot 3U and 24-slot 2U under 2,5" drives. They all include multiple 1Gb/s iSCSI ports for network connectivity, an architecture built with surveillance applications in mind that need fast connection to multiple clients. Up to 444 drives can be connected to expansion enclosures. With 10TB drive support, this means the available capacity can be up to 4PB.

    Series Composition EonStor DS 1000

    Models for 2.5" HDD

    DS 1024B - 2U, 24 2.5" drives with SAS or SATA interface

    DS-1036B - 3U, 36 2.5" drives with SAS or SATA interface

    Models for 3.5" HDD

    DS 1012 - 2U, 12 drives 3.5" with SAS or SATA interface

    DS 1016 - 3U, 16 drives 3.5" with SAS or SATA interface

    DS 1024 - 4U, 24 drive 3,5"with SAS or SATA interface

    Performance

    • EonStor DS 1000 provides up to 550K IOPS (cache operations) and 120K IOPS (full path, including disks) to speed up all storage-related operations.
    • Throughput reaches 5,500 MB/s read and 1,900 MB/s. on the record, which makes it easy handle even intensive workloads with high efficiency.

    Working with SSD Cache

    (optional, license required)

    • Improved read performance for hot data
    • Up to four SSDs per controller
    • Large pool capacity SSD drives: up to 3.2 TB

    Rice. 1 Growth in IOPS when SSD cache is saturated with hot data

    Combined host interface options

    • All systems feature four 1Gb/s iSCSI ports to provide more than enough connectivity to clients, servers, and other storage arrays.
    • Optionally added module host interface with 8 Gb/s or 16 Gb/s Fiber Channel, iSCSI 10 Gb/s or 40 Gb/s iSCSI, 10 Gb/s FCoE or 12 Gb/s SAS to run in parallel with default iSCSI ports 1 Gbps.
    • Optionally added to converged host board with 4 connectivity options to choose from (16Gb/s FC, 8Gb/s FC and 10Gb/s iSCSI SFP+, 10Gb/s FCoE)

    Various cache saving options

    Lifetime, maintenance-free, no-replacement supercapacitors and flash module provide a safe and reliable power source to maintain cache memory state if the main power supply fails

    Hot-swappable battery backup unit (BBU) with flash module stores data if the system suddenly shuts down or there is a power failure.

    You can choose BBUs or supercapacitors to suit your needs and budget

    Optionally available and included advanced features:

    Local replication Local Replication

    (Standard license is included by default, extended license is optional)

    Snapshots

    Standard license Extended license
    Snapshots per original volume 64 256
    Snapshots in the system 128 4096

    Volume Copy/Mirror

    Standard license Extended license
    Source volumes in the system 16 32
    Replication pairs per source volume 4 8
    Replication pairs per system 64 256

    Fine tuning (enabled by default)

    Just-in-time capacity allocation optimizes storage usage and eliminates dedicated but unused storage space.

    Remote replication (additional license)

    Replication per volume: 16
    Replication pairs per source volume: 4
    Replication batches per system: 64

    automated layered system data storage (additional license)

    Two or four storage tiers based on drive types

    SSD support

    Automatic data migration with scheduling options

    SSD caching (additional license)

    Accelerating data access in read-intensive environments such as OLTP

    Supports up to 4 SSDs per controller

    Recommended DIMM capacity per controller for SSD cache:

    DRAM: 2GB Max. SSD Cache Pool Size: 150 GB

    DRAM: 4 GB Max. SSD cache size: 400 GB

    DRAM: 8 GB Max. SSD Cache Pool Size: 800 GB

    DRAM: 16 GB Max. SSD cache size: 1,600 GB

    Doesn't fit your Infortrend DS 1000 Series storage system? Consider the storage of another series or line, go to the section:

    The humble DotHill 4824 storage system will be the hero of this review. Surely many of you have heard that DotHill, as an OEM partner, produces entry-level storage systems for Hewlett-Packard - those very popular HP MSA (Modular Storage Array) already in the fourth generation. The DotHill 4004 line matches the HP MSA2040 with minor differences which will be detailed below.

    DotHill is a classic entry-level storage solution. Form factor, 2U, two options for different drives and with a wide variety of host interfaces. Mirrored cache, two controllers, asymmetric active-active with ALUA. Last year, new functionality was added: disk pools with three-level tiering (tiered data storage) and SSD cache.

    Characteristics

    • Form factor: 2U 24x 2.5" or 12x 3.5"
    • Interfaces (per controller) 4524C/4534C - 4x SAS3 SFF-8644
    • Scaling: 192 2.5" drives or 96 3.5" drives supports up to 7 additional DAEs
    • RAID support: 0, 1, 3, 5, 6, 10, 50
    • Cache (per controller): 4GB with flash protection
    • Features: snapshots, volume cloning, asynchronous replication (except SAS), thin provisioning, SSD cache, 3-level tiering (SSD, 10/15k HDD, 7.2k HDD)
    • Configuration limits: 32 arrays (vDisk), up to 256 volumes per array, 1024 volumes per system
    • Management: CLI, Web interface, SMI-S support

    Disk pools in DotHill

    For those who are not familiar with the theory, it is worth talking about the principles of disk pools and tiered storage. More precisely, about a specific implementation in the DotHill storage system.

    Before the advent of pools, we had two limitations:

    • The maximum diskgroup size. RAID-10, 5 and 6 can have a maximum of 16 drives. RAID-50 - up to 32 disks. If you need a volume with a large number of spindles (for the sake of performance and / or volume), then you had to combine LUNs on the host side.
    • Suboptimal use of fast disks. You can create a large number of disk groups for several load profiles, but large numbers hosts and services on them, it becomes difficult to constantly monitor performance, volume and periodically make changes.

    A disk pool in DotHill storage is a collection of several disk groups with load distribution between them. In terms of performance, you can consider the pool as a RAID-0 of several subarrays, i.e. we are already solving the problem of short disk groups. In total, only two disk pools, A and B, are supported on the storage system, one per controller), each pool can have up to 16 disk groups. The main architectural difference is the maximum use of free placement of stripes on disks. Several technologies and features are based on this feature:

    Differences from HP MSA2040

    Performance

    Storage configuration
    • DotHill 4824 (2U, 24x2.5")
    • Firmware version: GL200R007 (latest at the time of testing)
    • Activated RealTier 2.0 license
    • Two controllers with CNC ports (FC/10GbE), 4 x 8Gb FC transceivers (installed in the first controller)
    • 20x 146GB 15Krpm SAS HDD (Seagate ST9146852SS)
    • 4x 400GB SSD (HGST HUSML4040ASS600)

    Host Configuration

    • Supermicro 1027R-WC1R platform
    • 2x Intel Xeon E5-2620v2
    • 8x 8GB DDR3 1600MHz ECC RDIMM
    • 480GB SSD Kingston E50
    • 2x Qlogic QLE2562 (2-port 8Gb FC HBA)
    • CentOS 7, fio 2.1.14
    The connection was made through one controller, direct, through 4 8Gb FC ports. Naturally, the mapping of volumes to the host was through 4 ports, and multipath was configured on the host.

    Pool with tier-1 and cache on SSD

    This test is a three-hour (180 cycles of 60 seconds) load with random access in 8KiB blocks (8 threads with a queue depth of 16 each) with various read / write ratios. The entire load is concentrated on the 0-20GB area, which is guaranteed to be less than the volume of the performance tier "and or the cache on the SSD (800GB) - this is done in order to quickly fill the cache or tier in an acceptable time.

    Before each test run, the volume was re-created (to clear the SSD-tier "a or SSD cache), filled with random data (sequential write in 1MiB blocks), read-ahead was turned off on the volume. IOPS, average and maximum latency values ​​were determined within each 60- second cycle.

    Tests with 100% read and 65/35 read + write were carried out both with SSD-tier (a disk group of 4x400GB SSD in RAID-10 was added to the pool), and with SSD cache (2x400GB SSD in RAID-0, storage does not allow more than two SSDs to be added to the cache for each pool.) The volume was created on a pool of two RAID-6 disk groups of 10 46GB 15K RPM SAS disks each (i.e. it is actually a 2x10 RAID-60). Why not 10 or 50?To deliberately make it harder for storage to write randomly.

    IOPS

    The results were quite predictable. As the vendor claims, the advantage of the SSD cache over the SSD-tier "ohm is faster cache filling, i.e. the storage system responds faster to the appearance of "hot" areas with an intense load on random access: IOPS grow on 100% reading together with a delay drop faster than in the case of using tier "ing.

    This advantage ends as soon as a significant write load is added. RAID-60, to put it mildly, is not very suitable for random writes in small blocks, but this configuration was chosen specifically to show the essence of the problem: the storage system cannot cope with writing, because. it bypasses the cache on a slow RAID-60, the queue fills up quickly, and there is little time left to service read requests even with caching. Some blocks still get there, but quickly become invalid, because the recording is in progress. This vicious circle causes the read-only cache to become inefficient under this load profile. Exactly the same situation could be observed with early versions of the SSD cache (before the advent of Write-Back) in LSI and Adaptec PCI-E RAID controllers. Solution - use a more productive volume initially, i.e. RAID-10 instead of 5/6/50/60 and/or SSD-tier instead of cache.

    Average delay


    Maximum delay

    This graph uses a logarithmic scale. In the case of 100% and using the SSD cache, you can see a more stable latency value - after the cache is full, the peak values ​​​​do not exceed 20ms.


    What can be summed up in the dilemma "caching vs. tiering"?
    What to choose?
    • Cache filling is faster. If your workload consists of predominantly random reads and at the same time the area of ​​"hot" periodically changes, then you should choose a cache.
    • Saving "fast" volume. If the "hot" data fits entirely in the cache, but not in the SSD-tier, then the cache will probably be more efficient. The SSD cache in DotHill 4004 is read-only, so a RAID-0 disk group is created for it. For example, having 4 SSDs of 400GB each, you can get 800GB of cache for each of the two pools (1600GB in total) or 2 times less when using tiering "and (800GB for one pool or 400GB for two). Of course, there is another 1200GB option in RAID-5 for one pool, if the second one does not need SSDs.

      On the other hand, the total useful pool size when using tiering will be larger due to the storage of only one copy of the blocks.

    • The cache has no performance impact on sequential access. When caching, blocks are not moved, only copied. With a suitable load profile (random reading in small blocks with repeated access to the same LBA), the storage system issues data from the SSD cache, if it is there, or from the HDD and copies it to the cache. When there is a serial access load, the data will be read from the HDD. Example: a pool of 20 10 or 15k HDDs can give about 2000MB / s with sequential reading, but if the necessary data ends up on a disk group from a pair of SSDs, then we will get about 800MB / s. Whether this is critical or not depends on the real scenario for using storage systems.

    4x SSD 400GB HGST HUSML4040ASS600 RAID-10

    The volume was tested on a linear disk group - RAID-10 of four 400GB SSDs. In this DotHill shipment, HGST HUSML4040ASS600 turned out to be abstract “400GB SFF SAS SSDs”. This is an SSD of the Ultrastar SSD400M series with a fairly high declared performance (56000/24000 IOPS for reading / writing 4KiB), and most importantly, a resource of 10 rewrites per day for 5 years. Of course, now HGST has more productive SSD800MM and SSD1600MM in its arsenal, but these are enough for DotHill 4004.

    We used tests designed for single SSDs - "IOPS Test" and "Latency Test" from the SNIA Solid State Storage Performance Test Specification Enterprise v1.1:

    • IOPS Test. The number of IOPS "s (IOPS) is measured for blocks of various sizes (1024KiB, 128KiB, 64KiB, 32KiB, 16KiB, 8KiB, 4KiB) and random access with different read / write ratios (100/0, 95/5, 65/35, 50/50, 35/65, 5/95, 0/100) 8 threads were used with a queue depth of 16.
    • Latency Test. The value of the average and maximum delay is measured for various block sizes (8KiB, 4KiB) and read/write ratios (100/0, 65/35, 0/100) with a minimum queue depth (1 thread with QD=1).
    The test consists of a series of measurements - 25 rounds of 60 seconds. Preload - Sequential write in 128KiB blocks until 2x capacity is reached. The steady state window (4 rounds) is verified by plotting. Steady State Criteria: Linear fit within the window must not exceed 90%/110% of the mean.

    SNIA PTS: IOPS test



    As expected, the declared performance limit of a single controller in terms of IOPS with small blocks was reached. For some reason, DotHill indicates 100,000 IOPS for reading, and HP for MSA2040 - more realistic 80,000 IOPS (40 thousand per controller is obtained), which we see on the graph.

    For verification, a single SSD HGST HGST HUSML4040ASS600 was tested with a connection to a SAS HBA. On a 4KiB block, about 50 thousand IOPS were received for reading and writing, with saturation (SNIA PTS Write Saturation Test) the write dropped to 25-26 thousand IOPS, which corresponds to the characteristics declared by HGST.

    SNIA PTS: Latency Test

    Average delay (ms):


    Max Delay (ms):


    Average and peak latency values ​​are only 20-30% higher than those for a single SSD when connected to a SAS HBA.

    Conclusion

    Of course, the article turned out to be somewhat chaotic and does not answer several important questions:
    • Comparison in a similar configuration with products from other vendors: IBM v3700, Dell PV MD3 (and other descendants of LSI CTS2600), Infrotrend ESDS 3000, etc. Storage systems come to us in different configurations and, as a rule, not for long - you need to load and / or deploy.
    • Storage limit not tested by bandwidth. We managed to see about 2100MiB/s (RAID-50 of 20 disks), but I did not test sequential load in detail due to the insufficient number of disks. I am sure that the declared 3200/2650 MB / s for reading / writing could be obtained.
    • There is no IOPS vs latency graph, useful in many cases, where, by varying the queue depth, you can see how many IOPS can be obtained with an acceptable latency value. Alas, there was not enough time.
    • Best practices. I didn't want to reinvent the wheel, because there is