Understanding Server Storage Technologies
As
the demand for server storage has grown, so too has the number of new
storage technologies. Over the years, the range of server storage
options has broadened from simple direct-attached storage (DAS) to
network-attached storage (NAS) and, most recently, to Fibre Channel
(FC) and iSCSI SANs.
Direct-Attached Storage
DAS
is storage attached to one server only. An example of a DAS solution is
a set of internal hard disks within a server or a rack-mounted RAID
that is connected to a server through a SCSI or FC controller. The main
feature of DAS is that it provides a single server with fast, block-based
data access to storage directly through an internal or external bus.
(Block-based, as opposed to file-based, means that data is moved in
unformatted blocks rather than in formatted files.) DAS is an
affordable solution for servers that need good performance and do not
need enormous amounts of storage. For example, DAS is often suitable
for infrastructure servers such as DNS, WINS and DHCP servers, and
domain controllers. File servers and Web servers can also run well on a
server with DAS.
The
main limitation of DAS is that it is directly accessible from a single
server only, which leads to inefficient storage management. For
example, Figure 1
shows a LAN in which all storage is attached directly to servers.
Despite the fact that the Web and App2 servers have excess storage,
there is no easy way for these resources to be redeployed to either the
Mail or App1 server, which need more storage space.
Managing DAS in Windows Server 2008
The
main tool used for managing DAS in Windows is the Disk Management
console. This tool, which you can access in Server Manager, enables you
to partition disks and format volume sets. You can also use the
Diskpart.exe command-line utility to perform the same functions
available in Disk Management and perform additional functions as well.
Network-Attached Storage
NAS
is self-contained storage that other servers and clients can easily
access over the network. A NAS device or appliance is a preconfigured
server that runs an operating system specifically designed for handling
file services. The main advantage of NAS is that it is simple to
implement and can provide a large amount of storage space to clients
and servers on a LAN. The downside of NAS is that, because your servers
and clients access a NAS device over the LAN as opposed to over a local
bus, access to data is slower and file-based as opposed to block-based.
NAS performance is, therefore, almost always slower than that of DAS.
Because
of its features and limitations, NAS is often a good fit for file
servers, Web servers, and other servers that don’t need extremely fast
access to data.
Figure 2 shows a network in which clients use a NAS appliance as a file server.
Managing NAS
NAS appliances come with their own management tools, which are typically Web based.
Storage-Area Networks
SANs
are high-performance networks dedicated to delivering block data
between servers and storage subsystems. From the point of view of the
operating system, SAN storage appears as if it were installed locally.
The most important characteristic that distinguishes a SAN from DAS is
that in a SAN, the storage is not restricted to one server but is, in
fact, available to any of a number of servers. (SAN storage can be
moved from server to server, but outside of clustered file system
environments, it is not accessible by more than one server at a time.)
Note: SAN vs. DAS
Although
DAS data transfer rates are typically faster than those of a SAN, the
performance gap between DAS and SAN technologies is constantly
shrinking. Despite the bus speed advantage offered by DAS, SANs are
still considered preferable because the advantage SANs offer of shared
storage outweighs the shortcoming of slightly lower access speeds.
A
SAN is made up of special devices, including HBAs on the host servers,
switches that help route storage traffic, disk storage subsystems, and
tape libraries. These hardware devices that connect servers and storage
in a SAN are called the SAN fabric.
All these devices are then interconnected by fiber or copper. Once
connected to the fabric, the available storage is divided up into
virtual partitions called logical unit numbers (LUNs), which then
appear to servers as local disks.
SANs
are designed to enable centralization of storage resources while
eliminating the distance and connectivity limitations posed by DAS. For
example, parallel SCSI bus architecture limits DAS to 16 devices at a
maximum (including the controller) distance of 25 meters. Fibre Channel
SANs extend this distance limitation to 10 km or more and enable an
essentially unlimited number of devices to attach to the network. These
advantages enable SANs to separate storage from individual servers and
to pool unlimited storage on a network where that storage can be shared.
SANs
are a good solution for servers that require fast access to very large
amounts of data (especially block-based data). Such servers can include
mail servers, backup servers, streaming media servers, application
servers, and database servers. The use of SANs also allows for
efficient long distance data replication, which is typically part of a
disaster recovery (DR) solution.
Figure 3 illustrates a simple SAN.
SANs generally occur in two varieties: Fibre Channel and iSCSI.
Fibre Channel SANs
Fibre
Channel (FC) delivers high-performance block input/output (I/O) to
storage devices. Based on serial SCSI, FC is the oldest and most widely
adopted SAN interconnect technology. Unlike parallel SCSI devices, FC
devices do not need to arbitrate (or contend) for a shared bus. FC
instead uses special switches to transmit information between multiple
servers and storage devices at the same time.
The
main advantage of FC is that it is the most widely implemented SAN
technology and has, at least until recently, offered the best
performance. The disadvantages of FC technology are the cost of its
hardware and the complexity of its implementation. Fibre Channel
network components include server HBAs, cabling, and switches. All
these components are specialized for FC, lack interoperability among
vendors, are relatively expensive, and require special expertise.
iSCSI SANs
Internet
SCSI (iSCSI) is an industry standard developed to enable transmission
of SCSI block commands over an Ethernet network by using the TCP/IP
protocol. Servers communicate with iSCSI devices through a locally
installed software agent known as an iSCSI initiator. The iSCSI initiator executes requests and receives responses from an iSCSI target,
which itself can be the end node storage device or an intermediary
device such as a switch. For iSCSI fabrics, the network also includes
one or more Internet Storage Name Service (iSNS) servers that, much
like DNS servers on a LAN, provide discoverability and zoning of SAN
resources.
By
relying on TCP/IP, iSCSI SANs take advantage of networking devices and
expertise that are widely available, a fact that makes iSCSI SANs
generally simpler and less expensive to implement than FC SANs.
Aside from lower cost and greater ease of implementation, other advantages of iSCSI over FC include:
Connectivity over long distances
Organizations distributed over wide areas might have a series of
unlinked “SAN islands” that the current FC connectivity limitation of
10 km cannot bridge. (There are new means of extending Fibre Channel
connectivity up to several hundred kilometers, but these methods are
both complex and costly.) In contrast, iSCSI can connect SANs in
distant offices by using in-place metropolitan area networks (MANs) and
(wide-area networks) WANs.
Built-in security
No security measures are built into the Fibre Channel protocol.
Instead, security is implemented primarily through limiting physical
access to the SAN. In contrast to FC, the Microsoft implementation of
the iSCSI protocol provides security for devices on the network by
using the Challenge Handshake Authentication Protocol (CHAP) for
authentication and the Internet Protocol security (IPSec) standard for
encryption. Because these methods of securing communications already
exist in Windows networks, they can be readily extended from LANs to
SANs.
Note: iSCSI SAN fabric
An
iSCSI SAN can use dedicated devices for its fabric, or it can rely on
an organization’s existing LAN, MAN, or WAN infrastructure. For both
security and performance, a dedicated iSCSI network separating network
traffic from storage traffic is recommended.
The
main disadvantage of an iSCSI SAN is that, unless it is built with
dedicated (and expensive)10-GB Ethernet cabling and switches, the I/O
transfer of iSCSI is slower than an FC-based SAN can deliver. And if
indeed you do choose to use 10-GB equipment for your iSCSI SAN instead
of the much more common choice of gigabit Ethernet, the high cost of
such a 10-GB solution would eliminate the price advantage of iSCSI
relative to FC.
Exam Tip
Vocabulary terms you should understand for the exam include LUNs, HBA, iSCSI initiator, iSCSI target, SAN fabric, and iSNS.
Managing SANs
Windows
Server 2008 includes the Virtual Disk service (VDS), an application
programming interface (API) that enables FC and iSCSI SAN hardware
vendors to expose disk subsystems and SAN hardware to administrative
tools in Windows. When vendor hardware includes the VDS hardware
provider, you can manage that hardware within Windows Server 2008 by
using tools such as Disk Management, Storage Manager for SANs (SMfS),
Storage Explorer, iSCSI Initiator, or the command-line tool
DiskRAID.exe.
SMfS
SMfS is available in Windows Server 2008 as a feature that you can add
through the Add Features Wizard. You can use SMfS to manage SANs by
provisioning disks, creating LUNs, and assigning LUNs to different
servers in the SAN.
Figure 4 shows the SMfS console.
Storage Explorer
Storage Explorer is available by default in Windows Server 2008 through
the Administrative Tools program group. You can use Storage Explorer to
display detailed information about servers connected to the SAN as well
as about fabric components such as HBAs, FC switches, and iSCSI
initiators and targets. You can also use Storage Explorer to perform
administrative tasks on an iSCSI fabric.
iSCSI Initiator
The iSCSI Initiator tool is available by default in Windows Server 2008
through the Administrative Tools program group. This tool enables you
to configure security, discovery, and other features of the local
server connections to iSCSI targets.
DiskRAID DiskRAID is a command-line tool that enables you to manage LUNs in a VDS-enabled hardware RAID.