1. The Search for Disaster Protection
Traditionally, large businesses have used a variety of
techniques to ensure that files stored on a server were both secure
and safe. These solutions tend to be expensive, but when spread across
all the supported workstations and buried in a large MIS budget they
are feasible. The same solutions would not be
feasible or acceptable in most small businesses, but that doesn’t
change our very real need to protect ourselves from disaster.
Fortunately, both hardware and software solutions can provide a very
high level of security and safety at a budget more in keeping with the
realities of a small business. However, before we talk about those
solutions, let’s make sure we all understand the terminology of
disk management. Let’s review some definitions:
Physical drive The actual hard disk itself, including the case,
electronics, platters, and all that stuff. It’s not terribly
important to the disk administrator.
Partition A portion of the
hard disk. In many cases, this is the entire hard disk space, but
it needn’t be.
Master Boot Record
(MBR) A technique for partitioning a hard disk. This is
the default method for Windows Small Business Server 2011.
MBR-partitioned disks are limited to a maximum of four partitions per disk, and a maximum size of 2
terabytes.
GUID Partition Table
(GPT) A technique for partitioning a hard disk,
GPT is replacing MBR for larger hard disks and large storage arrays.
Windows Small Business Server 2011 supports GPT-partitioned disks
for all disks except the boot disk. GPT disks support 128
partitions and are required for disks (or arrays)
larger than 2 terabytes.
Allocation unit The smallest unit of managed disk space on a hard disk or logical
volume—also called a cluster.
Primary partition A portion of the hard disk that’s been marked as a
potentially bootable logical drive by an operating system. MS-DOS
could support only a single primary partition, but Windows Server 2008 can
support four primary partitions on an MBR hard disk and 128 primary
partitions on a GPT hard disk.
Extended partition A nonbootable portion of the hard disk that can be
subdivided into logical drives. There can be only a single
extended partition per hard disk, but this partition
can be divided into multiple logical drives. Extended partitions
are deprecated in Windows Small Business Server 2011 and can’t be
directly created from the GUI.
Volume A unit of disk space
composed of one or more sections of one or more dynamic
disks.
Simple volume The dynamic equivalent of a partition. A portion of
a single dynamic disk, a simple volume can be assigned either a single drive
letter or no drive letter and can be attached (mounted) on zero or
more mount points.
Extended volume Similar to, and sometimes synonymous with, a
spanned volume, an extended volume is any dynamic
volume that has been extended to make it larger than its original
size. When an extended volume uses portions of more than one
physical disk, it is more properly referred to as a spanned volume.
Logical drive A section or partition of a hard disk that acts as
a single unit. An extended partition can be divided, for example,
into multiple logical drives.
Logical volume Another name
for a logical drive.
Basic disk A traditional disk
drive that is divided into one or more partitions, with a logical
drive in each primary partition. Basic disks do not support the more advanced
functions of disk management, but they can be converted to
dynamic disks in many cases.
Dynamic disk A managed hard
disk that can be used to create various volumes.
iSCSI (Internet Small Computer Systems
Interface) A protocol for using remote, centralized, storage
as if it were local. Uses either shared or dedicated TCP/IP
networks. Traditionally, they were reserved for storage area
networks (SANs) with specialized (and very expensive) hardware,
but now they’re easily available with software
implementations.
iSCSI target The iSCSI server or provider.
iSCSI initiator The client or requester for an iSCSI storage
device.
LUN (Logical Unit
Number) The “disk” that an iSCSI target presents to an
iSCSI initiator. A LUN can be any portion of the available storage
on the iSCSI server.
RAID (redundant array of independent
[formerly “inexpensive”] disks) The use of multiple hard disks in an array to
provide for larger volume size, fault tolerance, and increased
performance. RAID comes in different levels, such as RAID-0, RAID-1, and RAID-5. Higher numbers Don’t
necessarily indicate greater performance or fault tolerance, just
different methods of doing the job.
Spanned volume A collection
of portions of hard disks combined into a single addressable unit.
A spanned volume is formatted like a single drive and can have a
drive letter assigned to it, but it will span multiple physical
drives. A spanned volume—occasionally referred to as an
extended volume—provides no fault tolerance
and increases your exposure to failure but does permit you to make
more efficient use of the available hard disk space.
Striped volume Like a spanned
volume, a striped volume combines multiple hard disk portions into
a single entity. A striped volume uses special formatting to write
to each of the portions equally in a stripe to increase
performance. A striped volume provides no fault tolerance and
actually increases your exposure to failure, but it is faster than
either a spanned volume or a single drive. A stripe set is often
referred to as RAID-0, although this is a misnomer because plain
striping includes no redundancy.
Mirror volume A pair of dynamic volumes that contain identical
data and appear to the world as a single entity. Disk mirroring
can use two drives on the same hard disk controller or use
separate controllers, in which case it is sometimes referred to as
duplexing. In case of failure on the part of
either drive, the other hard disk can be split off so that it
continues to provide complete access to the data stored on the
drive, providing a high degree of fault tolerance. This technique
is called RAID-1.
RAID-5 volume Like a striped volume, this combines portions of
multiple hard disks into a single entity with data written across
all portions equally. However, it also writes parity information
for each stripe onto a different portion, providing the ability to
recover in the case of a single drive failure. A RAID-5 volume provides excellent throughput for read
operations but is substantially slower than all other available
options for write operations.
SLED (single large expensive
disk) Now rarely used, this strategy is the opposite of
the RAID strategy. Rather than using several inexpensive hard
disks and providing fault tolerance through redundancy, you buy
the best hard disk you can and bet your entire network on it. If
this doesn’t sound like a good idea to you, you’re right. It’s
not.
JBOD Just a bunch of disks. The hardware equivalent of a
spanned volume, this has all the failings of any spanning scheme.
The failure of any one disk will result in catastrophic data
failure.
Note:
Additional RAID levels are supported by many hardware
manufacturers of RAID controllers. These include RAID 0+1, RAID-10,
RAID-6, and RAID-50. For more details on various RAID levels, see
the manufacturer of your RAID controller or http://en.wikipedia.org/wiki/RAID#Standard_levels.
There were basically three possible
technologies available: Modified Field Modification (MFM), Pulse Frequency Modulation (PFM), and Small Computer System (or Serial) Interface
(SCSI). Unless you were a total geek (and had oodles of
money), your systems used either MFM or PFM, and RAID wasn’t even an
option. Over time, SCSI became the only real choice for the vast
majority of servers and even became mainstream on high-end
workstations. Servers at the high end might use fiber, but SCSI had
the vast majority of the server disk market.
Integrated Device Electronics (IDE), later called Advanced
Technology Attachment (ATA), became the standard on the personal
computer. However, IDE never made serious inroads into the server
market because, although it was fast for single tasks, it lacked the
inherent multitasking support and bus mastering that a server disk
interface technology required, and there were no real hardware RAID
solutions that supported it. Largely supplanted by Serial ATA
(SATA) even on personal computers, this technology has
no place at all on your server.
The introduction of SATA technology has made serious inroads
into the lower end of the server marketplace. With SATA RAID controllers built into many motherboards,
and stand-alone SATA RAID boards that support eight or more SATA
drives and have substantial battery-backed RAM cache onboard, many
low-range to mid-range servers are finding SATA RAID solutions to
provide a cost-effective alternative to SCSI. While most SATA RAID
controllers lack the ability to hot-swap a failed drive, and
generally don’t have the performance potential of SCSI or Serially
Attached SCSI (SAS), they are still quite attractive alternatives
where cost is a primary factor. SATA also makes sense as secondary
or “near-line” storage for a server.
The new kid on the block, however, is SAS. This is the most
interesting addition to the server storage equation in quite a
while. Using the same thin cables and connectors as SATA, with none
of the configuration nuisance of traditional SCSI, SAS is definitely
the way to go. When combined with new 2.5-inch drives, the ability
to put a really large amount of very fast storage in a small space
has taken a significant step forward. Many SAS controllers fully
support SATA drives also, allowing you to combine the two
technologies on the same controller.
With the main bottleneck for servers continuing to be I/O in
general, and especially disk I/O, there will continue to be pressure
to find new and faster methods to access disk-based storage. Using
wide arrays of fast, traditional disks—especially using low-power,
high-density 2.5″ SAS disks—enables fast and flexible storage arrays
in remarkably smaller spaces, and with lower energy and cooling
requirements.
A new option that directly addresses the limitations of
traditional spinning disk technologies is the solid state drive
(SSD), a “disk” that is actually a collection of flash
memory that connects to a SATA controller. SSDs are currently still
quite expensive and not really ideal for large RAID arrays because
of performance degradation over time, but the technology is rapidly
improving and offers promise for the future.
2. Choosing the Storage Solution for Your Network
The first decision you need to make when planning your storage
solution for SBS is really made when you specify your server. If your
budget can afford it, you should definitely consider choosing a
hardware RAID solution that lets you add disks on the fly and
reconfigure the array without turning off the server or rebooting.
This is absolutely the best and most flexible storage solution for
protecting your data, and it can take the form of hot-swappable SAS
hard drives, or even a SAN. The best choices aren’t cheap, and in most
cases you need to make at least some portion of the decision as part
of the original server purchase.
|
Although most hardware storage solutions require you to make
decisions very early in the buying process, a growing number of
Network Attached Storage (NAS) solutions can provide a cost-effective way to
increase the storage flexibility of your SBS network. Many of the
available solutions, especially at the lower end of the price range,
are designed more for home networks and digital media sharing than
for business networks. However, there are also excellent NAS servers available that are based on Microsoft
Windows Storage Server. These provide the greatest flexibility and
support for an SBS network, and we prefer them when adding a NAS to
an SBS network because they also support iSCSI protocols. For more
on Windows Storage Server–powered NAS servers, see http://www.microsoft.com/windowsserver2008/en/us/wss08.aspx.
Another interesting option is
Windows Storage Server 2008 R2 Essentials (WSSE). Although it’s not designed primarily as a NAS,
WSSE supports much of the same functionality, while adding in the
ability to do client backups very efficiently. If your need for a
NAS is primarily to add some near-line storage for occasional-use
files, or to store local backups, we think WSSE is a very
interesting alternative, and one we’re using on our personal SBS
network.
|
After the server is actually in place and is being used, you
can’t really make a change to the underlying hardware for your
existing storage that would allow you to use a hardware RAID
solution—at least not easily. But you can add a
hardware RAID controller and a RAID array when it’s time to add more
storage to the server, and you can also use the built-in facilities of
SBS to make your existing disk subsystem more fault-tolerant by using
dynamic disks and the software RAID of SBS.
2.1. Storage Connection Technologies
Your choices are
Integrated Device Electronics
(IDE) Strictly a client solution. It’s inexpensive, but
not appropriate on a server. It’s now being replaced even at the
client end by SATA.
Serial Advanced Technology
Attachment (SATA) A newer and faster version of IDE.
It’s still primarily a workstation solution, but it’s acceptable
when combined with hardware RAID for smaller servers.
External Serial Advanced Technology
Attachment (eSATA) A way to use SATA for external, secondary, or
backup storage.
Small Computer System Interface
(SCSI) Perfect for servers and high-end workstations,
but significantly more expensive than SATA. It has the ability
to have up to 13 drives per SCSI channel.
Serially Attached SCSI
(SAS) Perfect for servers. This is a relatively new
technology that is rapidly becoming the mainstream server
storage interface. Prices are still more than SATA.
Internet SCSI
(iSCSI) Important for SANs, and can even be used as a
boot device for SBS servers. Hardware SANs are generally well
outside the budgets of most SBS networks, but software SANs
based on Windows Storage Server or third-party iSCSI software
are a very viable option for adding storage flexibility to your
SBS network.
FireWire Hot-pluggable. This is a good choice to use as a
backup storage device.
Universal Serial Bus
(USB) Only appropriate if you use USB 2.0 or later.
It’s good for CD and DVD drives, and it’s hot-pluggable. It’s
also a good choice for use as a backup storage device,
especially USB 3.0.
Fibre Channel A great option if you have really large amounts
of money to spend.
Network Attached Storage
(NAS) A good way to provide large amounts of storage
that can be flexible to meet your needs. Specify Windows Storage
Server–based NAS for the greatest flexibility and
compatibility.
Storage Area Networks
(SAN) Faster and more flexible than the typical NAS,
but also much more expensive and difficult to configure.
Hardware SANs are generally not for small business networks, but
software-based SANs are becoming a viable option.
Solid State Disks
(SSD) Initially used primarily for notebook computers,
these are starting to find their way into servers—especially
high-density servers in data centers, where their power savings
are a plus. They’re still too pricy for most SBS
networks.
