IP Storage: What Lies Beyond

IP storage infrastructure meets technical
requirements of traditional SAN at reduced performance levels.

IP storage is a new technology area that is fast evolving to become an
industry standard with increasing support from industry leaders like Microsoft,
Cisco, and EMC. IP storage leverages on Internet SCSI (iSCSI) protocol for
networking the storage devices and host systems. iSCSI encapsulates SCSI
commands into IP packets and transmits them over IP networks.

This
means that standard IP networks can be used for building storage networks to
connect storage devices and host servers that technically meet the requirements
of a fibre channel storage area network (SAN). These IP SANs leverage standard
network infrastructure components.

Types of storage

NAS: NAS connects to the local area network (LAN) infrastructure through
Gigabit, Fast Ethernet or ATM technologies. NAS enables client systems on the
network to access storage at the file level using remote file access protocols
such as Network File System (NFS) and Common Internet File System (CIFS).

FC SAN: From the connectivity perspective, fibre channel is flexible almost
like a general communication network and has low protocol overheads similar to
SCSI or ESCON. Fibre channel supports transporting protocols such as IP, HIPPI
or SCSI etc.

In a FC SAN, fibre channel transports SCSI over the network connecting hosts
and storage devices. Host servers connecting storage devices therefore will
access storage at SCSI block level.

Arun  Rao

Over the last decade both NAS and FC SAN technologies have matured and now
support a wide range of operating systems and applications. However, both these
technologies serve two different technological requirements and remain
complementary to each other. FC SAN is definitely an expensive infrastructure
that requires fibre channel host bus adapters, cabling, and switches in addition
to the storage devices themselves.

Flexibility, low cost, availability of higher bandwidth, and ability to
transport different types of data have made IP networks an attractive option for
building storage networks. iSCSI supported storage devices allow hosts to access
storage resources at block level as in a traditional FC SAN.

From the hardware connectivity perspective both NAS and iSCSI resemble each
other. Both storage devices connect to the IP network. The difference is in the
level at which the host systems access the storage — file level in case of NAS
and block level in case of iSCSI.

It is important to note that unlike NAS and FC SAN, there are only a limited
number of operating systems that support iSCSI. Hence, while looking at iSCSI as
a cost effective option against FC SAN, support for various operating systems in
the environment needs to be assessed.

IP storage Deployment scenarios

Storage on IP is possible in several permutations.

a) Simple IP SAN: A conceptual layout of iSCSI storage network is
shown in the figure. Host systems are connected to the IP storage device using
standard local area network infrastructure.

In figure 1.1, it can be noted that application performance will be greatly
affected due to the protocol overhead inherent to IP networks. The host bus
adapters connecting host servers to IP storage or the host operating system
needs to take care about encapsulation of SCSI commands to transmit through IP
networks.

Network adapters with TCP off-load engine ensures that the operating systems
are not loaded with the responsibility of protocol conversions. In IP SAN,
overall performance can be improved using host bus adapters with TCP off-load
engines.

IP Storage is typically used in consolidating application servers that
require low performance block level access to the storage device.

 b) IP SAN and FC SAN on common fibre channel storage: As shown
in figure 1.2, iSCSI can also be used in requirements where fibre channel SAN
and IP SAN needs to co-exist. The figure illustrates building Fibre Channel SAN
and IP SAN on a single Fibre Channel Storage System. In order to consolidate IP
SAN and FC SAN on a single fibre channel storage system, FC to IP gateways are
used as indicated in the figure 1.2.

 c) Data replication over WAN link: Another deployment of iSCSI
is to perform data replication from one fibre channel storage system to another
as a means for disaster recovery (as shown in figure 1.3). When the disaster
recovery site is thousands of miles away from the primary site, using fibre
channel links for data replication across storage systems is not feasible. iSCSI
helps using WAN infrastructure to perform block level replication of data across
the two locations. Source volumes from primary storage system can be replicated
to target volumes in a secondary storage system over WAN links as indicated in
figure 1.3 below.

 IP storage can be ideal in software houses for application development
and testing especially in database environments. Database environments will
typically require direct attached storage for hosting databases.

 IP storage can be deployed in research & development organisations
to build large IP SANs cost-effectively. The primary storage infrastructure
requirement of research organisations and development houses is to satisfy their
technical needs and performance is not a major concern. IP storage
infrastructure meets technical requirements of traditional SAN at reduced
performance levels.

Summary

iSCSI as a storage networking technology will evolve further to support a
variety of operating systems and applications. There will be a performance
difference between traditional fibre channel SAN and IP SAN while both these
technologies will continue to scale to support higher bandwidths. Storage
networking will use all the three key technologies — NAS, FC SAN and IP SAN to
support requirements of various applications and also to optimise the storage
networking cost.

Arun Rao is Head-Storage Solutions at Datacraft India.