Fujitsu ETERNUS DX 100 S3 storage system review and test

Today we will talk about Fujitsu's entry-level scalable storage system, the DX 100 S3. Moreover, the emphasis on scalability and unification is not without purpose. The fact is that this system can be expanded to high-end models by simply replacing controller blocks and can provide both block (via Fiber Channel, FCoE, iSCSI, SAS and InfiniBand) and file access to data (NFS), and without any whatsoever additional devices (filers). Although it should be noted that there is an opportunity to save: if file access is not required, you can purchase a version, in the factory configuration of which there will be no fee in the controller unit responsible for file access.

This system belongs to the updated ETERNUS S3 line, which received the SAS3 interface (12 Gbit SAS), which doubled its throughput. And the VxWorks OS (developed by Wind River Systems), on which the storage system is running, is a multi-threaded, 64-bit real-time operating system providing high performance and security. A similar operating system is used in the NASA space program, in particular, under its control, the Curiosity spacecraft was landed on the surface of Mars.

Key features and benefits of the ETERNUS DX 100 S3 storage system:

• Performance and capacity reserve providing data aggregation for server, email, database and commercial applications virtualization in one system
• Maintain high data volume growth at low cost and investment protection due to the high degree of extensibility within the system and the ability to transition to more powerful systems
• The included ETERNUS SF Express Management software package with a wide range of functions.
• Disaster recovery flexibility
• Built-in intelligent snapshot backup features
• Convenient and economical operation with minimal administrative effort
• The use of a unified system of access to disk drives SAN and NAS, contributing to quick payback
• Reduced operating costs due to the unification of management functions in all systems of this family
• A wide range and various combinations of network types are supported, as well as a direct connection to the server
• Support for various combinations of 2.5 and 3.5 inch discs

')
Specifications:

• Interfaces - Fiber Channel (up to 16 Gbit / s), FCoE (10 Gbit / s), iSCSI (10/1 Gbit / s), SAS (12/6/3 Gbit / s), CIFS, NFS
• Disk devices (maximum) - 144 x SSD, SAS and Nearline SAS in mixed configurations (3.5 inches / 2.5 inches)
• Capacity (maximum) - 576 TB
• Number of host interfaces - 4
• Number of connectable host ports - max. 1024
• RAID level - 0, 1, 1 + 0, 5, 5 + 0, 6

Test system configuration:

• Fujitsu ETERNUS DX 100S3 storage with block access, switching over FC 8 Gbit channel.
• 10 SAS 900 GB / 10K rpm drives - RAID 10
• Fujitsu PRIMERGY RX200 S8 Server, Intel Xeon E5-2660, 8x8Gb RAM, 2x QLogic FC HBA
• ESXi 5.5 + vm Windows Server 2012R2 + LUN 100GB

From the point of view of the physical performance of the device there is nothing to complain about, but it is also impossible to isolate anything special. ETERNUS DX 100 S3 is a good-quality system made of high-quality materials, with a design already established among all manufacturers, dictated by rack-mount design.

The graphical management interface of the storage system is ascetic and does not stand out in any particular way, these are all the same top tabs with horizontal tree sub-items and parameter selection tables. The weakest side of this interface is the implementation of performance monitoring, it seems to be there, but it is rather present “for show”. The first thing that strains is that monitoring must be turned on necessarily, otherwise the readings will be even zero, and its inclusion occurs in a completely different menu, so you will not find it right away. But even after switching on, we get a non-configurable list of indications, without any graphs or visualization, but with the possibility of saving the log for a certain period of time. But all the ugliness of the GUI is compensated, without exaggeration, by the number of settings outstanding in this segment of data storage systems. For example, flexible mapping configuration, both by ports and by port-groups, LUN-groups and hosts. It is possible to select the mode of operation with the host, namely: we can choose which scheme will use asymmetric access to the ALUA (Asymmetric Logical Unit Access) logical drive: Active-Active, Active-Passive or Preferred Path. We can also specify how much cache will be available to a particular logical drive and more.







Based on the manufacturer’s claims that this system is a full-fledged Active-Active device, one can expect data processing in parallel by two controller blocks, regardless of the controller “owner” of the logical disk, which was confirmed when testing the system. Selecting the ALUA Active-Active mode on the storage side and applying the Round Robin multipassing policy in the ESXi settings for the disk presented to the DX 100 S3 hypervisor, we received two active channels and load balancing on both controllers. This can be seen on the ESXi monitoring graph and the indications of a uniform CPU load on both storage controllers when recording the test file with the IOmeter program.



On the graph you can see that the total speed of the recorded data is close to 500,000 kilobytes per second, passing through adapters (FC HBA - vmhba 1 and 2) connected to different controllers, 250 000 Kb / s.
Below are the average test results for various utilities running in the Windows Server 2012 virtual OS deployed in the VMware ESXi 5.5 virtualization environment.

AnvilBenchmark








These test results are not straightforward, the Anvil and CrystalDiskMark utilities have reflected very good performance indicators for this segment, and IOmeter, on the contrary, is nothing outstanding. This is due to the fact that the standard synthetic tests Anvil and CrystalDiskMark have specified parameters intended more for testing drives (SSD, HHD) than storage systems, and reflect mainly the speed of the disk subsystem taking into account the fast cache (in our case, 8 GB 4 GB per controller): the more cache and disks, the more I / O operations we receive. IOmeter allows you to more flexibly load the system, overflow the cache, and increase the queue of requests. And to achieve maximum results, close to the real load, you need to know exactly what we want from the storage system - how many IOPS (Input / Output Operations Per Second) we need (for this we collect the required number of disks, otherwise the system simply will not have them take) what percentage will be read and what is written, how many approximately simultaneous requests will come (queue depth) and so on. And here the possibility of flexible configuration of Fujitsu ETERNUS DX will come in handy. But even if you use it "as is", with the default settings, we will get the maximum of opportunities, multiplied by the modern reliable software and hardware.

Summing up, it is worth noting two main factors that form an opinion about the Fujitsu ETERNUS DX 100 S3. The first is the ease of commissioning (no mistakes, problems or difficulties with setting up and connecting, everything is clear and everything works), the second is the flexibility of the system (this is a lot of settings, and the ability to get both block and file access, as well as upgrade to more productive storage systems by simply replacing controllers).

Prepared on the material of Tretyakov Vyacheslav, an engineer of the company "Paradigma". See the full article here .