| Surviving
the data explosionChoosing the right storage solution for your network neednt be an
either/or choice between a SAN (Storage Area Network) or NAS (Network Attached Storage).
In fact if you think the two formats are mutually opposed youve got it all
backwards, so to speak. Geoff Marshall offers some advice on getting the best from both
storage worlds
Last month we
looked at the new concept of the Storage Area Network (SAN), which is often confused with
Network-Attached Storage (NAS). Here we're going to explain the differences, focussing
more on NAS.
The Data Explosion
By the year 2000, the estimated annual growth in demand for online data storage will be
running at over 50 Petabytes (1 Petabyte = 1,000 Terabytes) with PC, UNIX and NT users
generating over 90% of this. Companies are realising what an enormous data management
headache this presents. Consequently, enterprise data storage is now a necessity as
corporate information is moved to the strategic centre of technology operations. However,
data is no longer restricted to a single host machine, but has moved to an open system,
often in mixed-platform environments.
This proliferation of open systems, often involving many servers, PCs, and laptops has led
to data being distributed across heterogeneous computer systems and networks. Business
data objects no longer reside within single data architectures and consist of multiple
information structures such as records, fields, text, images, or audio. Consequently, a
data storage (or archiving) system must be able to support all data types stored on
multiple platforms. Don't lose that data!
Data distribution has also created other problems, particularly in the areas of security,
access and availability. Disaster recovery procedures and resilient data back-up routines
are vital for organisations that depend on their electronic data for doing business -- we
all know that any data loss could have catastrophic consequences, especially if the loss
is irretrievable. This leads to the storage medium. Data backups and archives must be
robust, safe and secure because users have a habit of wanting their data back at some
point!
The future of tape looks assured -- or does it? Tape life is finite and users know they
must backup their tape backups at least every two years before the medium starts to
degrade. However, the short life of tape does mean that future sales of tape cassettes are
guaranteed -to committed tape users. Likewise manufacturers are encouraged to continually
develop tape drive technology. The logistics of storing tapes is labour intensive -
changing cassettes, labelling, tape libraries, catalogues, physically moving tapes to
safe, secure sites and the time to find and restore data can run into days!
NAS or SAN ?
NAS defines an open architecture for managing and consolidating data storage operations in
today's corporate and enterprise networks. They are designed
to accelerate application performance, giving users shared data access in heterogeneous
environments, as well as enabling administrators to consolidate
storage from multiple desktop and server systems into a centrally-managed pool of data.
One of the issues surrounding SAN (described fully last month) is it's shortcomings in
these areas:
Data to desktop
Heterogeneous data sharing
Lack of standards for file access
and locking
The best of both worlds?
One company, Network Appliance, has combined the benefits of SAN and NAS (Network-Attached
Storage) to produce a complete solution that gets away from such shortcomings. The new
combined SAN and NAS architecture supports increased enterprise scalability and utilises
existing industry-standards like TCP/IP and Fibre channel. Data sharing, storage
management, backup and network administration become simpler and more effective, leading
to reductions in total cost of ownership.
The latest report from IDC predicts the growth of the NAS market in excess of 55% per year
through 2003. This IDC report also ranks Network Appliance number one in NAS revenue with
41% of the overall market share for 1998. We spoke to the UK Managing Director of Network
Appliance, Steve Ronksley, who says, "The explosion of data in the work place has
generated a strong demand for storage architectures that can solve the business problems
of managing network accessible data. As the market leader in network attached storage,
Network Appliance is focused on providing dedicated filers that deliver on our reliable,
fast and simple appliance philosophy. The company's philosophy is to design specialised
servers, called appliances, which are dedicated to particular data access and storage
tasks, thus optimising performance, ease of deployment and manageability. These appliances
deliver fast, simple, reliable, and cost-effective access to data stored in the
network."
Dan Warmenhoven, the President and Chief Executive Officer of Network Appliance, believes
that there is an increasing amount of confusion over the type of storage markets:
"The industry is not about Network Attached Storage verses Storage Attached Networks
- it's NAS plus SAN. SAN encompasses NAS."
Decisions, Decisions
If you're still confused about the difference between SAN and NAS, here's some more
questions (and answers) that are often asked by network managers: Do I go for a SAN based
solution, or for a NAS device? Is NAS a stop gap to SAN? Will there still be a need for
NAS devices when SAN's mature? Will the same NAS device support my need for file serving
as well as database application serving? How do I protect my data once it's on the NAS?
These are just a few of the issues that face those considering these emerging storage
technologies.
The Storage Area Network adds a new transport layer to the traditional client/server model
between the application server and the storage. The benefits of separating an application
between the client and the server with a LAN has been well demonstrated over the last two
decades, but this model is starting to show the signs of ageing under the weight of the
current explosion in data storage.
The SAN gives the adopter the ability to do traditional storage management tasks such as
backup and restore, clustering and replication without impacting the LAN or any of it's
servers, and also opens the way for true device and data sharing on a scale previously
thought impossible. Hence SANs are, in the short to medium term, thought to offer a
server-to-server, server-to-storage or storage-to-storage solution. Put more simply, they
are aimed at the back-end datacenter. Clients will still run applications across the LAN,
with the application servers accessing their data across the SAN.
A NAS or Network Attached Storage device however, combines traditional disk array
technology with intelligence. This intelligence is achieved by adding a small
processing unit into the same case as the disk, which is commonly based on an Intel or
SPARC RISC chip. In this way, the disk array can be attached directly to an Ethernet based
network. We are all familiar with the idea of a network printer - a printer that anyone
with the correct rights in the network can print to. The NAS device follows the same
principle. It is a disk that is shared out over the Local Area Network using conventional
file sharing protocols such as NFS or CIFS, to all the users with appropriate rights
within the network. All the users in the network are able to mount filesystems directly
from the NAS across the network without having to route the data via a server.
NAS devices raise a new set of challenges for storage management vendors. For example, how
do you integrate a NAS device, or even a farm of NAS devices into our traditional backup
and restore plan? If NAS devices are not directly attached to any server, how are we able
to transfer data from the NAS to the backup server which has the tape storage sub-system
attached? NAS devices tend to store in the region of hundreds of gigabytes, through to
tens of terabytes of information, meaning that even if you have just invested in the
latest gigabit Ethernet network, it is a huge amount of data to shovel across the LAN to a
backup server. Having off-loaded the disk-based storage from your server, wouldn't it be a
good idea to also move some of the relevant tape-based storage?
We spoke to Jonathan Martin, EMEA Product Marketing Manager at Veritas Software
Corporation, about this NAS problem: "Veritas has taken a pivotal role over the past
two years in the definition of a new industry standard protocol, which concerns itself
with just this issue. NDMP, or the Network Data Management Protocol, which allows you to
connect a tape library directly to the back of the NAS device, and drive the management of
this tape library from your Veritas NetBackup server. By sending a new set of NDMP control
commands from the Veritas NetBackup server to an NDMP-compliant NAS device, data is moved
via SCSI from the NAS disk to the tape library, while the metadata (what got backed up,
where it got backed up to, etc) is stored on your Veritas NetBackup server as usual. In
this, way, you are backing up large amount of consolidated data, without impacting the
production servers that are using that data.
"The level of intelligence that is built into the current crop of NAS devices
wouldn't notch up anything Einsteinian in an IQ test, but they do give you the basic
ability to consolidate data into a central storage resource, and make it available to a
wide number of users. Veritas is working with a number of it's strategic vendors on a
second generation of NAS devices, something that Veritas is branding an iNAS (or
intelligent NAS) device. It is only a small step to make from running networking software
on the NAS CPU, to running other applications. In this instance, we are talking about a
stripped down set of the classic Veritas storage management technologies such as Veritas
File System, Veritas Volume Manager and Veritas NetBackup. The management of the storage
is therefore no longer driven from the server, but is housed on the storage itself freeing
up the ever more precious CPU time on server for its applications.
Conclusion
Both SAN and NAS devices have a valid part to play in the evolving storage market, and
require a different approach to managing them. See http://www.ndmp.org for more details of
the Network Data Management Protocol. |

Steve Ronskley: A
strong demand for storage architectures that can provide network accessible data

Above : The F760 Tower
from Network Appliance

Jonathan Martin: NAS
CPUs can be made to run applications other than networking software
|