Which of the following is dedicated server device designed solely for providing shared storage for network users?

2.5.2 Storage virtualization

Virtualization provides fault tolerance and failover features by automatically failing over to active nodes. Due to hardware abstraction, it is also easy to establish a DR environment with relatively little cost and effort. These attributes of virtualization help in high availability. We have looked at various aspects of virtualization in the previous chapter. Storage virtualization can be adopted to provide a highly available storage system.

Storage virtualization can be achieved using SAN, which is discussed in the next section.

NAS

Network-Attached Storage (NAS) is a computer data storage connected to a network, providing data access to various group of clients. NAS not only operates as a file server, but it is also specialized for this task either by its hardware, software, or configuration of those elements.

NAS systems are networked appliances which contain one or more hard drives, often arranged into logical, redundant storage containers, or RAID. Network-attached storage removes the responsibility of file serving from other servers on the network. They typically provide access to files using network file sharing protocols. In CCTV, NAS devices have gained popularity, as a convenient method of sharing files among multiple computers. The benefits of network-attached storage, compared to file servers, include faster data access, easier administration, and simple configuration.

Although it may technically be possible to run other software on a NAS unit, it is not designed to be a general purpose server. For example, NAS units usually do not have a keyboard or display, and are controlled and configured over the network, often using a browser. A full-featured operating system is not needed on a NAS device, so often a stripped-down operating system is used, like FreeNAS, a simplified version of FreeBSD.

12.3.3.2 Elements of a NAS device

Figure 12.35 shows some components of a NAS device. Actually, there are more components that should be included, which are discussed below:

12.3.3.2.2 Storage elements

A disk controller provides RAID options and LUN (see Figure 12.37).

Figure 12.37. NAS device hardware elements.

12.3.3.2.3 Software elements and file system

Software elements include the NAS OS software, NIC microcode, and RAID software. The NAS OS consists of four components (see Figure 12.38):

Figure 12.38. NAS device software elements.

•

Micro-kernel OS

•

Memory management

•

Resource management

•

Cache management

The file system is the most important element for NAS. It is an open standard or protocol to support file sharing (see Figure 12.39).

Figure 12.39. NAS file system.

To implement a NAS solution, we have two options: one is an integrated solution and the other is a gateway one. With an integrated NAS device or solution, all NAS hardware and software components are integrated into one physical frame or enclosure. It becomes a self-contained environment.

12.3.3.2.4 Integrated NAS

An integrated NAS solution means that the NAS device serves storage clients via the IP network. It can be either a lower cost solution via an ATA type of interface connected to a single storage enclosure or a high-end solution with an FC interface connected to storage arrays.

A low cost integrated NAS solution is very common for departmental NAS applications because its primary focus is to consolidate many storage devices. In order to reduce opex, the flexibility and management for a NAS device’s configuration is kept to a minimum. In other words, the solution is fixed. The device cannot be upgraded beyond its initial configuration. When any extra capacity is required, the solution just connects additional new boxes to the IP network (see Figure 12.40).

Figure 12.40. A low-end application-type of integrated NAS solution.

For a high-end integrated NAS solution, we can add an external and dedicated NAS device or storage array. Because it is the integrated NAS device that is preconfigured, it has limited scalability.

III.A.4. Network Attached Storage (NAS) Systems

Network Attached Storage (NAS) systems are another file service device. The NAS is connected to the LAN just like a file server. Rather than containing a full-blown OS, it typically uses a slim microkernel specialized for handling only I/O requests such as NFS (UNIX), GIFS/8MB (Windows 2000/NT), and NCP (NetWare). Adding or removing a NAS system is like adding or removing any network node. For example, it doesn't get much simpler than the Meridian Data Snap server—containing only an on/off switch and an ethernet port. It provides an instant storage boost by simply plugging it into the network switch or hub port. However, the NAS is subject to the variable behavior and overhead of the network, which makes NAS less desirable in many cases.

7.1.6 Home Clouds

Network Attached Storage (NAS) devices allow individuals and small businesses especially to have their own private cloud data storage. Security concerns will motivate many to provide their own cloud data storage solutions, both private and commercial. Physical control of both endpoints, devices and storage, does satisfy some security concerns, related to legal issues and insider access. However, it does place the burden of security on the owner of home/private cloud. In addition it does not address data backup and storage location redundancy.

Search engines like Shodan (http://www.shodan.io/) crawl the Internet specifically looking for devices connected to it, including NAS devices. Banners returned by these devices upon connection often reveal or hint at vulnerabilities. A security firm recently developed a proof-of-concept worm that infects and propagates via NAS devices. Cyber criminals have already compromised NAS devices to mine Bitcoins, steal data, and encrypt data to hold for ransom (Constantin, 2014).

NAS Devices

NAS devices are appliances with the sole purpose of providing data storage. It can be challenging to obtain a forensic image from a NAS device since they run limited services and protocols. If you can acquire the image forensically through an attached system, that may be the preferred option. Otherwise, you may need to disassemble the NAS device and image it drive by drive. Many NAS devices are designed and marketed for the home or small business user. They are no longer just in the realm of enterprises.

So, how do we follow the traditional best practices again when there is no practical way to access the drives directly and take physical images? The other very real consideration with large storage systems is that the necessary hardware requires a large investment. Therefore, it would be logical to assume that the system is attached to a system that is at least marginally important. For a business that needs its systems running to generate revenue, it may again become a business decision to limit the scope of work to limit the downtime.

Network attached storage

NAS has been traditionally used by small to medium businesses within their enterprise networks. This is sometimes called file storage as the applications deal with storage at the file or directory level instead of at the block level, which is referred to as block storage in storage area networks. Figure 8.4 shows how NAS arrays could be connected to an enterprise Ethernet LAN to provide storage to various clients.

These storage arrays are similar to the SAN storage arrays described above but have additional processing resources to convert file level data into block level data for storage on the individual disk drives. Later in this chapter, we will look inside a storage array and provide more details on both SAN and NAS applications.

NAS is intended to be a low-cost way to add shared storage for multiple users in the network, but it is not as reliable or secure as a dedicated SAN, and performance is lower due to the fact that storage data is shared with other data on the same Ethernet network. This convergence of storage and data on a single network has been addressed in layer 3 TCP/IP networks using the iSCSI protocol and in layer 2 Ethernet networks through Fibre Channel over Ethernet (FCoE) which was standardized in 2010. Both will be discussed later in this chapter.

17.4.3 Network Attached Storage

Network attached storage (NAS) is a common component of supercomputing installations. It provides centralized shared storage capability, frequently with very large capacity, to multiple hosts, specifically including compute and login nodes. While SANs provide shared access to mass storage at the device level, NAS operates at file-system level. The accessing clients use specifically designed libraries or kernel extensions to import data volumes hosted by NAS servers. Remote data shares may be mounted on the client side to provide practically identical application programming interfaces to those exposed by local file systems, such as Portable Operating System Interface I/O. The contents of remote data shares may be then accessed using standard utilities and libraries that have been developed to support “regular” files, effectively making the fact that the communication with the server and the data transfer are performed over the network completely transparent to the application.

NAS implementations utilize a handful of network file system protocols. The commonly used ones include Server Message Block (SMB), originally developed by IBM and Microsoft, Common Internet File System (CIFS), which is a more feature-rich version of SMB, Apple Filing Protocol (AFP), a proprietary protocol used by Apple File Service, and Network File System (NFS), which originated at Sun Microsystems. While the first two are usually found in Microsoft DOS and Windows-based environments, AFP is restricted to Apple products and NFS is broadly employed in the Unix world, including Linux. NFS is an open standard defined in the Internet Engineering Task Force/Internet Society Request for Comments and has open-source implementations. SMB functionality is available on Unix-compatible platforms thanks to the open-source SMB/CIFS reimplementation known as Samba. Finally, AFP is supported by the open source Netatalk project. All these protocols rely on TCP/IP for connectivity, although some SMB and NFS variants are capable of datagram-based communication (User Datagram Protocol).

A high performance NAS server, depicted in Fig. 17.23, derives from the architecture of a conventional compute node. The primary differences are possible inclusion of multiple network adapters to provide the necessary data bandwidth to clients and a substantially expanded storage pool. The latter usually requires multiple controller boards to provide the required number of ports for connecting the storage devices and optionally to incorporate hardware-level data protection, such as RAID. The server should have a sufficiently large memory pool to be able to accommodate a large number of outstanding I/O requests and efficiently buffer data. Due to the increased power draw caused by the large storage pool, a NAS server should also be equipped with redundant power supplies of appropriate rating and make allowance for sufficient case ventilation to evacuate the generated heat. Since a single server will eventually hit a performance barrier, a clustered NAS may be considered to enable capacity scaling. A clustered NAS takes advantage of distributed (Ceph, AFS, GFS, and others) or parallel (GPFS, Lustre, PANFS, OrangeFS, PVFS, and others) file systems to provide an abstraction of a single logical file system comprising all storage devices while enabling high-bandwidth access to file data and load distribution across the component servers.

Networked Storage—NAS Filer

Network-attached storage (NAS) is in some ways easier to accelerate than SAN, and there are some low-hanging fruit in the tuning game. Any drive in an NAS box can be an SSD and provide acceleration. I would advise any CIO to make sure that the CEOs NAS share is on SSD!

Making a filer all-SSD may not work as well as expected, since the controller board or network ports may be too slow. At a minimum, the controller board (usually a COTS motherboard) should have two 10 GbE ports if SSD are deployed, while four ports would be even better if more than six SSD are in the same box. It’s just a matter of bandwidth. In 2016, 25 GbE will start to enter mainstream and if this is in your datacenter master plan, from 2017 on, any upgrade of NAS networking and any new NAS boxes should have two or four 25-GbE ports [21].

The higher throughput of SSD may need some tuning on the controller, too. Slow motherboards (We often use hand-me-down servers for filers) won’t cut it with a bunch of SSD, while DRAM may need some expansion as caches grow.

ARM-based filers [22] much like the Lego appliances talked about earlier are also a possible solution. Most datacenters already use multiple filers for a variety of reasons, so the Lego appliance fits in. With object stores (especially those based on Ceph) getting a filer interface, this might be a good time to transition to the new technology.

More complex filer configurations, such as NetApp F-series units, handle tougher challenges such as virtual desktop or virtual image booting. Since most of the data and apps in a desktop are identical over many users, logic suggests that a cloning/deduplication system would work well. If the virtual image is stored in a fast flash unit, such as a PCIe card, access can be very rapid indeed. NetApp has a cloning tool that allows images to be torn down and then rebuilt with new files in just seconds.

While the NetApp approach [23] can work with huge sets of images, it still transmits all of the data to the servers. That means the network is still a bottleneck. The solution of course is not to transmit the images and that has led us to the containers approach to virtualization. Here, the image (OS + tools + selected apps) is sent just once to each server, reducing both the network load and the bandwidth passing through the NAS controller by as much as 300×, or even much more.

Since the common image isn’t tenant-dependent, it is possible to store it on a local small SSD [24] in each server, speeding boot even further and removing common image traffic altogether from the storage network. This will make container boot much faster than hypervisor network boot and likely will help push the industry toward containers.

AFA units haven’t had anywhere near the impact on the NAS side of networked storage. Most filers don’t share data with other filers, so having a sharable repository like an AFA doesn’t work. Making the AFA into a filer is a possibility, and Violin has a product that aims at SMB3, but even there the bottleneck of a single filer engine handling all the file systems and protocol slows things down.

In sum, filers do benefit from SSD, but the limitations of motherboards in the controller and networks tend to limit the number of drives. AFAs are of little utility today, though this may change. Flash accelerators can, however, have huge impacts in specific use cases.

Storage Area Networking (SAN) Switches

Specialized switches called SAN switches are at the heart of a typical SAN. Switches provide capabilities to match the number of host SAN connections to the number of connections provided by the storage array. Switches also provide path redundancy in the event of a path failure from host server to switch or to switch or from storage array to switch. SAN switches can connect both servers and storage devices and thus provide the connection points for the fabric of the SAN. Sometimes modular switches are interconnected to create a fault-tolerant fabric. For larger SAN fabrics, director-class switches provide a larger port capacity (64–128 ports per switch) and built-in fault tolerance. The type of SAN switch, its design features, and its port capacity all contribute to its overall capacity, performance, and fault tolerance. The number of switches, types of switches, and manner in which the switches are interconnected define the topology of the fabric.