IDC Adaptec iSCSI Manual

WHITE PAPER
iSCSI: Networked Storage Without the Pain
Sponsored by: Adaptec
Robert C. Gray
January 2004
INTRODUCTION
iSCSI builds on the 25-year experience curve and maturity of Ethernet and IP
networking. iSCSI permits companies to treat server I/O as just another application on
the message/data network. Potentially (and in the most favorable case) it allows
companies to gain the benefits of storage area networks (SANs) by applying unused
portions of IP/Ethernet switches. IT staffs require minimal training beyond their
current IP proficiency to gain SAN competency. Moreover, IP-based SANs mean the
possibility of remote block storage from any location with a sufficient IP connection.
If you are paying increasing attention to your server storage, then it is probably time
to begin contemplating an external storage consolidation strategy. In the past, such a
strategy frequently started with the provisioning of dedicated network attached
storage (NAS) file servers. With the advent of iSCSI technology, there are stronger
values and return on investment (ROI) from starting with an iSCSI-based SAN. For
this document, IDC spoke with IT managers who are directly responsible for the
selection and deployment of iSCSI within their organizations. The case studies in this
white paper demonstrate how iSCSI fulfilled companies' requirements and met their
new challenges.
ISCSI MARKET SUMMARY
IDC interviewed more than 1,400 organizations as the iSCSI specification approval
process was nearing conclusion, but before suppliers could put iSCSI products in
users hands. The research indicates a pent-up demand among users for the benefits
of iSCSI. Based on this primary research, IDC developed a detailed worldwide
forecast for iSCSI adoption (see Figure 1). In 2003 and 2004, growth is throttled by a
limited product set of storage target arrays from suppliers. When that bottleneck
breaks in late 2004, IDC expects a rapid growth in the purchase of iSCSI products.
Global Headquarters: 5 Speen Street Framingham, MA 01701 USA P.508.872.8200 F.508.935.4015 www.idc.com

2 #3885 ©2004 IDC
FIGURE 1
Worldwide iSCSI-Based Disk Array Sales in SAN and DAS
Environments, 2003–2007
0 500 1,000 1,500 2,000 2,500 3,000 3,500
2007
2006
2005
2004
2003
($M)
Source: IDC, 2004
ISCSI: A DISRUPTIVE TECHNOLOGY, AND
WHAT IT MEANS TO USERS
Periodically a new way of doing something is so different, yet improved, that it
changes everything. IDC, in following Clayton Christensen's definition in his book The
Innovator's Dilemma, refers to such a technology as a disruptive technology. In the
past, mini–steel mills disrupted the integrated steel producers and hydraulic
excavators disrupted cable steam shovels. Disruption begins, typically, when basic
users are overserved by the expensive products of the era. The disrupting technology
is initially good enough for the overserved basic users and is inevitably much
cheaper. Over time the disrupting technology improves and displaces the incumbent
products. IDC, in partnership with Christensen's company Innosight, has declared
iSCSI a disruptive technology.
Figure 2 shows that IP storage networking improves over time with intelligent (TCP
offload engine [TOE]) adapters, with improved RDMA protocols, and, finally, with
speed capabilities of 10Gb, 40Gb, and 100Gb.

©2004 IDC #3885 3
FIGURE 2
IP/Ethernet as a Disruptive Technology
Source: IDC, 2004
Today, iSCSI complements Fibre Channel; it is ideal for situations in which Fibre
Channel is excessive for basic needs.
WHEN DIRECT ATTACHED SERVER STORAGE
BECOMES A PROBLEM, ISCSI SANS ARE THE
SOLUTION
The iSCSI SAN easily expands to incorporate NAS file serving. However, the
opposite is not always true: You cannot always easily expand your NAS file server to
accommodate block I/O server traffic. The easy-to-install and easy-to-use fixed NAS
proposition is exactly what gets in the way when you need the scalability and flexibility
of a SAN architecture to meet multiple years of changing and expanding needs.
Some issues that IT executives are experiencing today include:
! CIOs and IT executives are under the heavy pressure to deliver more with less.
! After years of growing IT budgets, many IT executives are faced with flat or even
declining resources. The problem is not growing staff but increasingly looking for
ways to apply technology to substitute for reduced staff.
High end
SAN
Midrange
Entry
Direct connect
Fibre Channel Ethernet/IP
Time
10Gb
RDMA
VI
TCP offload
iSCSI
Requirements
High end
SAN
Midrange
Entry
Direct connect
Fibre Channel Ethernet/IP
Time
10Gb
RDMA
VI
TCP offload
iSCSI
Requirements

4 #3885 ©2004 IDC
! The days of purchasing technology for technology's sake are over. Meeting
business objectives is goal number one. Support for continuous global business
processes with complex internal and external dependencies is the norm.
! Capacity continues to grow. Business operations generate new data daily.
Business analysis requires that more data be kept online and for longer periods
of time. Annual capacity growth of 30–100% is common.
! Server proliferation has increased as prices have dropped below capital expense
minimums and business units and work teams develop applications to support
their missions.
! Bladed servers promise to explode the complexity of the datacenter. Diskless
blades with networked storage are the most efficient architecture, especially
when data sharing is involved.
! Finally, even seemingly mundane applications such as email are now mission
critical, increasingly demanding redundant systems and rapid recovery capability.
Growth Inhibitors
Why users delay the adoption of SANs
We asked the following question of more than 100 end users in small companies
(fewer than 100 employees) and medium-sized organizations (100–999 employees):
Which of the following are your reasons for not having a SAN? The reasons are listed
in order of most frequent mention:
! #1 — Direct attached storage (DAS) is adequate. DAS increasingly becomes a
problem as the number of instances increases. Upgrades and backup become
increasingly time consuming, and compromises in quality of service (QoS) and in
recovery are inevitable.
! #2 — SANs are too expensive. iSCSI, combined with capacity-optimized disk
drives and improved management functions, will greatly reduce the cost of SANs.
! #3 — Don't want a special network. Supporting a Fibre Channel–based SAN
involves incorporating all new network skills. Protocol, hardware, suppliers, and
management tools are all new and different. iSCSI utilizes the IP/Ethernet
products already deployed. Employee skills are transferable.
! #4 — Don't need performance/availability. Many Windows servers and some
Unix servers seldom reach gigabit traffic levels. The corresponding applications,
while critical, achieve availability through server redundancy or rapid repair.
iSCSI/IP networking routing is sufficient to deal with congestion or failed
components.
In summary, respondents cited the high costs and lack of benefits associated with current
Fibre Channel–based technology as inhibitors to adoption. Moreover, the integration of a
new networking technology (Fibre Channel) was a barrier. Interestingly, the vast majority
of large organizations already have SANs. We found very few that did not.

©2004 IDC #3885 5
IDC expects iSCSI to cost less, feature more benefits, and use familiar Ethernet as
the interconnecting media. Clearly, the rollout of iSCSI will accelerate SAN adoption
by bringing SAN benefits to a complementary set of applications and organizations.
We also asked what benefits drew those already using SAN to adopt the technology.
The following bullets show the most important reasons for shifting from DAS to
networked SAN storage.
Why Users Deploy SANs
One hundred and seventy users responded to the following question: Which of the
following reasons for installing or evaluating a SAN is most important? The following list
consolidates similar reasons into fewer generic categories. The reasons are listed in
order of most frequent mention. It is not a surprise that "simplification" was the dominant
motivator of IT managers. One of the most enduring "dances" among organizational
groups is the tendency of business units (profit centers) to proliferate computer
applications and equipment on behalf of their mission. Likewise, the tendency of IT
organizations is to look to consolidation on behalf of their partnership with business
units in providing outstanding service while containing or even reducing costs.
! #1 — Reduce storage complexity. The need to do more with less is the most
frequent driver of SAN investments. Users indicated that reducing storage complexity
includes bringing storage to one location; standardizing on suppliers; reducing the
number of physical units to manage; and gaining simplified storage management.
! #2 — Save money. Users expected to save operational costs and to reduce
capital invested by eliminating stranded capacity.
! #3 — Improve backup. Networks allow a single backup strategy and uniform
and consistent data protection. SAN-connected arrays often offer a snapshot
data copy facility that allows backup to occur concurrently with ongoing data use.
The results are no backup window and frequently no need for staff off-prime and
staff overtime for copy to tape. Backup traditionally means that data stored on
disk drive is copied to tape for protection.
! #4 — Improve performance. SANs offer high-speed paths between servers and
storage for fast response and rapid data transfer.
! #5 — Better data management. Data sharing and data consolidation were
mentioned as motivators.
! #6 —– Better availability. SANs provide redundant data paths between servers
and storage and support redundant server clusters and RAID arrays with failover
capability. The net is near continuous data availability.
What changes with infrastructure complexity? The need for a structure around
infrastructure management becomes imperative. For smaller datacenters, additional
attention to process and staff training may temporarily handle growing complexity.
Eventually, IT executives are driven toward a consolidation strategy, wherein
automation and fewer points of management with less frequent and simpler

6 #3885 ©2004 IDC
involvement are possible. A number of signals indicate that consolidated network
storage is the right choice for your situation, including:
! Backup consistency. Too many of the steps in backing up DAS are subject to human
error. As the number of units to backup increases, typically multiple tools are involved
and staffs are forced to go to the server and its directly connected backup devices.
Fortunately, most IT executives seldom have to demonstrate that a backup really exists
and the server can be restored. Horror stories about the other guy abound.
! Backup in a shrinking window. Distribution of staff and computer integration
with suppliers and customers are happening even in smaller organizations. As
data increases in gigabytes, the window to back it up shrinks. It staffs are forced
to do the most critical backup work in the more disadvantaged time slots. The
best of your staff members work hard to avoid these assignments.
! Hassle of changes. With DAS, you may know you need more disk space only
when you have run out. The server may already be full of disk drives. Or the
storage is fine, but it is server upgrade time. IT staffs must manually and
physically respond to every new need, which occurs as an unscheduled
interruption to otherwise scheduled work. Dealing with these changes becomes a
QoS issue with internal clients and a morale issue with IT staffs.
! Need for availability. Most organizations are totally dependent on the proper
operation of their computer and storage infrastructure. An outage causes various
levels of pain and real dollar loss. With the cost of hardware having decreased
dramatically and pain increasing, companies can financially justify an investment in
redundancy for specific applications and business processes. The losses from lack of
continuous operation exceed the cost of duplicate servers, data paths, and disk parity.
! Need for shared storage. Something as simple as an organization's contact list
illustrates the need for data sharing because it is the customer list, the supplier
list, and the prospect list. Unless an organization shares this information, it risks
data duplication, work replication, and errors of omission and inclusion.
Computer sharing works a whole lot better than manual processes.
! Stranded capacity. We have listed last the failure to be able to redeploy unused
disk capacity or to have a high portion of installed capacity used. The rapid increases
in storage capacity per dollar, which have averaged about 40% per year
compounded, greatly compensate for the wasted invested capital of unused capacity.
We find in our research the first five pain points are the senior business issues.
It can no longer be a surprise that SANs represent the fastest-growing storage
investment. SANs have heretofore been architected around specialty networking
technology. The original SANs were created for mainframes with a now-slow 20MBps
ESCON interconnect. By the mid-1990s, in a move to bring the benefits of shared-
network storage to UNIX systems that were rapidly taking on mission-critical
applications, the storage industry invented Fibre Channel SANs. Tens of thousands of
Fibre Channel–based SANs have proven the concept of shared, network storage to
deliver on the value of simplified storage management and improved and more
uniform backup to high-end open systems environments.

©2004 IDC #3885 7
As the century turned, datacenter server proliferation exploded in less expensive, Windows-
based and Linux servers. The pain of managing these growing server pools calls for a more
efficient SAN technology for the masses of industry-standard servers. IDC believes the
answer mainly lies in the newly approved iSCSI protocol standard. iSCSI brings nearly all
the Fibre Channel SAN benefits to a lower cost, complementary sector of the IT datacenter.
iSCSI SANs will bring the availability and scaling characteristics, the data sharing, and the
manageability characteristics to a complementary set of applications.
PROFILE OF ADAPTEC ISCSI PLATFORM
The iSCSI storage application protocol is the core enabling technology foundation for
a major shift in the storage systems market. Many have said that iSCSI is such a
large change that it will truly disrupt the present market. We believe that the following
technologies will converge with the iSCSI protocol to fuel the disruption:
! TCP offload engines (TOEs). This breakthrough moves much of the computational
overhead of the TCP/IP software stack into hardware. Thus it largely eliminates the
heavy burden the application CPU normally has when TCP traffic is high. Although
intelligent TOE host bus adapters/network interface cards (HBA/NICs) will add to
cost, IDC believes they will be needed on only the high-end portion of the market.
! Capacity-optimized disks (e.g., SATA). In recent years, the relaxed
performance environment of the desktop has resulted in the industry's largest
capacity disk drives being offered with the ATA (and soon Serial ATA [SATA])
interface for the desktop. Recently, the price per gigabyte gap has increased four
times or more in favor of SATA disks. We believe that iSCSI target arrays, using
RAID and cache to provide availability and performance for ATA disks, will prove
a very competitive offering for capacity-oriented applications.
! Network storage processors. Network storage processors vary in function, but they
are essentially a RAID controller on a chip. We see these processors applied to provide
very cost-competitive RAID functionality to SATA disks in coming generations.
As these innovations are incorporated into storage systems, they will raise the
capabilities of iSCSI products and make them effective and suitable for an expanding
set of needs and applications. During the same adoption cycle, we expect suppliers to
offer NAS heads/gateways for inexpensively providing file views from iSCSI SANs
wherever needed. Finally, the Ethernet road map to 10Gb, 40Gb, and 100Gb speeds
will provide performance growth.
CASE STUDIES OF ISCSI USE
Major MidWest University
Where do you turn when you're a university with rapidly growing unified messaging
(email) and exploding streaming video capacity; 50,000 users; and the realization that
both applications are mission critical with five-nines availability needs?
You look at SANs. But what do you do when you do not have a million dollars for a
conventional SAN project? A major Midwest university and its consultants placed an early
bet on iSCSI.

8 #3885 ©2004 IDC
Purchase price as well as seamless integration and support drove the initial choice for
iSCSI. High availability, simple management, and high performance have been the
achieved bonuses. The developers compute that their iSCSI architecture was nearly an
order of magnitude less expensive to deploy at the 10TB level. The operations staff
assimilated the iSCSI storage with no staff increases and actually came to prefer it to
the embedded DAS products that iSCSI replaced. The iSCSI storage looked mostly like
just another IP network application. iSCSI's stability has been a surprise for these
pioneering early adopters. "It's been like a router, installed and then not touched for
years," says the IT consultant. "We were hoping to get near five-nines availability, but so
far we have 100% uptime for the past year." As for performance, the new iSCSI-based
storage system outperformed the parallel SCSI DAS systems it was replacing. The
parallel SCSI was Adaptec's not-too-shabby 2940. "We were really pleased with
Adaptec's 7211 iSCSI adapter. We had anticipated the applications' CPU offload, but
we were surprised by the positive impact it had on applications' performance."
iSCSI has also begun to change the way users think about and use storage at this
university. It has allowed many applications to shift to a more robust clustered architecture.
Previously, no applications were clustered because of the high cost of external networked
storage. iSCSI has also allowed many user groups at the university to gain the benefits of
centralized external storage, dedicated to their needs. iSCSI is so convenient that IT is
increasingly called upon to provide yet another "routable HDD."
The university and its conference peers are so committed to IP-based networked
storage that they are now architecting a network upgrade with 40Gb and 10Gb speeds
with the assumption that iSCSI will be a major traffic load. "FC [Fibre Channel] will not
go away; it's a lot like the transition from mainframes to Unix — those who had
mainframes did not walk away from them, but the expansion and new applications were
enabled by the newer technology," says the consultant. "My takeaway to those reading
this is for them to know that iSCSI is the storage solution right now."
Trimble
Trimble, a leading provider of positioning technologies, is in the process of moving to
a converged storage infrastructure to support an increasingly global, mobile, and
decentralized workforce. The company has four main sites in the United States —
Sunnyvale, California, its headquarters; Dayton, Ohio; Westminster, Colorado; and
Chandler, Arizona — but Trimble is truly a global organization. The company's 2,200
employees work across a total of 40 sites in more than 20 countries worldwide.
Trimble is looking to technologies such as iSCSI storage to gain broad SAN benefits
without the cost or hassle of dedicated storage network technologies.
Maximizing every employee's productivity is crucial, and with only a small IT staff
worldwide, which is distributed among two major and several minor sites, their
effectiveness is even more critical. IT management's goal is to use new technology to
deliver better functionality. Delivering better functionality today means rapid adoption
of new technologies with rapid training/learning. "This allows us to tackle real
problems like higher uptime through rapid recovery and periodic on-demand
capacity supporting engineering project teams with half a terabyte and more space for
chip simulations. The final consideration that makes iSCSI the best choice for
engineering is block storage over Gig-E. Speed and time to market are essential

©2004 IDC #3885 9
to staying on top. Everything we can do to enhance our engineers' ability to make or
exceed their time to market objectives is what we must do.
"We have gained system efficiencies using Linux and are taking advantage of the
networking big-time price drops of Gig-E. We needed the corresponding technology
leverage in storage. iSCSI is it.
"My staff had no fear of iSCSI. There was no special equipment to install and no thick
books to read. It looked like, smelled like, and tasted like SCSI. There just happened
to be a Gig-E switch in between. They told me this stuff is easy. The Adaptec HBA
was a champ. It was easy to install and configure, it made sense, and [it] worked
perfectly the first time and every day since then."
Trimble will first use iSCSI to support engineering projects with terabytes of simulation
space networked to desktop workstations around a six-building campus. The second
phase will bring a new level of robustness to its unified messaging system with an
iSCSI-based mirroring system for rapid recovery of Microsoft Exchange. "We have
very large exchange databases, and a disaster recovery that takes 7 to 8 hours is
not good for business. iSCSI will fix that." The third deployment of iSCSI will add
sharable capacity for the Oracle 11i test and QA systems. Currently, these needs are
supported with direct attached storage, which is usually 80% empty.
Expected iSCSI business benefits are as follows:
! Faster time to market with new product designs
! More REVs to its Oracle-based business process systems
! Email system supported as the mission-critical application it is
"My experience with iSCSI has been great — zero failures during our evaluation. I like
it because we can creep into the technology and evolve. We don't have to convert.
We are not about maximum storage performance. We are about maximizing our IT
staff resources and delivering new value to both engineering and business users."
END-USER CHALLENGES
In shifting from DAS to iSCSI-based storage, end users face several challenges.
However, a little forethought and consideration will help companies successfully
overcome the challenges, which leads to more mature IT organizations.
! The big step from DAS. Networked storage in SANs has been a complex,
delicate infrastructure. Thus, SANs have been the province of very large
organizations with project budgets in the hundreds of thousands of dollars. iSCSI
and IP-based SANs change that landscape and bring SAN benefits to smaller
organizations and departments that were previously cut off from the benefits.
However, this change means that companies must put attention on storage
needs and separate storage investments from server purchases.

10 #3885 ©2004 IDC
! ROI analysis and computation. Separate investments in storage infrastructure
demand an increased awareness of the total costs of storage management and
the costs of potential outages. The ability to articulate these total costs is
important in justifying the initial iSCSI investments.
! Process and policy: achieving the benefits. The best equipment in the world will
provide poor service without proper policies and practices. Successful storage
management is a combination of both. Unified networking and IP-based storage
promises a simpler, consistent environment and the ability for even smaller IT
organizations to attain a high level of competency and consistency of execution.
CONCLUSION/ESSENTIAL GUIDANCE
We find end users are widely aware of iSCSI technology and expect to adopt it as
products become available. Thus, we offer the following guidance:
! Be an early adopter of iSCSI/Ethernet external storage arrays. By taking this first
step, you are buying SAN-ready storage and will accumulate rapid experience
and confidence in iSCSI products. Make iSCSI your default DAS purchase.
! Keep initial iSCSI SANs on separate, dedicated networks. Do not mix LAN and
SAN traffic. Do not expose your server I/O to either LAN security issues or to
compromised congestion-based performance degradations.
! Build and use your initial SAN as support to a specific application, gaining
availability and data sharing.
! Use and treat iSCSI as a complement to Fibre Channel. If you have a Fibre Channel
SAN, consider if gateways from Fibre Channel to iSCSI are an appropriate method of
extending your servers supported by SAN storage. Use iSCSI for SANs where the
cost and complexity of Fibre Channel have been inhibitors. Use iSCSI where Fibre
Channel overserves the needs of the applications.
! Use ROI and the projected operational savings as well as initial capital savings to
build a powerful business case for the iSCSI deployment plan.
Copyright Notice
External Publication of IDC Information and Data — Any IDC information that is to be
used in advertising, press releases, or promotional materials requires prior written
approval from the appropriate IDC Vice President or Country Manager. A draft of the
proposed document should accompany any such request. IDC reserves the right to
deny approval of external usage for any reason.
Copyright 2004 IDC. Reproduction without written permission is completely forbidden.