Cover Story: Storage Management
Selecting the right storage platform reduces TCO
A well-designed network storage architecture
helps control storage expenditure and reduce TCO while managing explosive growth
in data volumes, and ensuring high-availability of data to end-users say Akhtar
Pasha and Abhinav Singh
Texas Instruments India saw its data triple
from 3 TB in December 2000 to 9 TB in 2001. Today the company sits on a mountain
of data—40 TB in all. Ensuring that this data is available to TI's chip designers
all of the time is a formidable task. Kameswar V Nagavarapu, General Manager,
IT, Texas Instruments (India) Pvt Ltd says, "Designing and planning networked
storage is very critical and is the 'life line' of the company." He adds, TI's
chip designing activity depends on data availability, which in turn depends
upon careful planning, designing and deployment of scalable storage solutions.
In an enterprise set-up designing and planning
the selection and deployment of networked storage is critical. Based on the
company's business goals and future plans, with regard to expansion and growth,
enterprise storage needs to be scalable. Organizations need to be in a position
to optimize their storage resources and bring down the TCO. The ability to recover
data in case of any eventuality—planned or unplanned—and policies for maintaining
business continuity are other key components of any storage architecture.
|"Ultimately NAS and SAN will merge. An
enterprise SAN over Fiber Channel provides dedicated point-to-point connections
between servers and disks without any IP contention" — Vinod Sadarangani,
Vice President-Information Technology, Orange
Professional Services Growing in Popularity
Designing network storage can be taxing;
it is a complex task. Companies are often assisted in the task by the professional
services divisions of storage vendors or integrators. Consultants such as Movinture
Storage Networks help organizations design and implement networked storage.
Consulting firms offer design, planning and implementation services.
Design is considered a full service rather
than an ad-hoc pre-sale quick fix. The focus lies more on integrating boxes,
both existing and those freshly purchased.
N. Ravishanker, assistant vice president
(IT), Bank of Muscat says, "After our merger with Centurion Bank, we will make
an RFP for building an enterprise storage network across the organization. At
that time we plan to hire the services of a neutral consulting firm for designing
our network storage architecture. We want someone who can evaluate and suggest
relevant products, and implement the same. We don't plan to take the services
of storage hardware vendors because their solutions tend to be biased."
Adds Prashant Prakash, founder and chief
strategy officer of Netkraft Pvt Ltd, "Being an IT Services company, storage
is one area that's not our core-competency. Taking the services of a storage
vendor will result in getting tied down with that vendor's products. For this
reason, we have chosen Movinture Storage Networks, as they are neutral and provide
a holistic view of our storage infrastructure required for business continuity
and disaster recovery."
That said, many companies such as HDFC
bank, TI, iGATE prefer to design networked storage architecture in-house. P.
S. Karthikeyan, the chief delivery officer (CDO) at iGATE Global Solutions Ltd
(IGS was formerly known as Mascot Systems) says, "We prefer to design our own
networked storage architecture with inputs from subject experts. IGS is a BS7799
certified Company and has taken a conscious decision not to outsource any activities
that are directly related to our business activities and information security."
Kameswar adds, "Being a global semiconductor
design company we would like to design our storage architecture ourselves. We
leveraged our IS team's storage and networking skill sets and know what best
fits our requirement." He adds the spec prepared by TI is given to the vendor
and the company tests their products itself.
Smaller companies tend to go with the storage
hardware vendor for designing and implementing networked storage. "A lot of
organizations don't have the expertise in-house," says Arun Rawtani, country
TSG manager, EMC Information systems NV. "Designing and implementing storage
requires a combination of networking and storage skills."
What's the best storage platform?
Karthikeyan says, "The general criteria
that we would use to pick the right storage solution is to identify the problem
that needs to be solved and analyze which solution will provide the greatest
return on investment (ROI). This would include factors such as capacity, performance,
scalability, manageability, and dependability, and last but not the least, the
Madhusadan Rao B, business manager, Apara
Enterprise Solutions (P) Ltd. says, "It's important to select a storage solution
that's independent of any hardware platform that you are working on, such as
HP UX or Solaris. The solution should be scalable and if the enterprise shifts
to any other platform tomorrow, the storage solution should be compatible with
that platform as well."
Surajit Sen, regional manager west, Network
Appliance adds that the most important thing is that organizations must ensure
that they base their architecture on open standards that are flexible, and can
accommodate all the protocols (TCP/IP or Fiber channel).
Designing storage network architecture,
be it NAS or SAN, depends upon a number of factors. The general practice at
mid- and large enterprises has been to outsource their network storage design
and planning to the storage hardware vendor.
S. R. Balasubramanian, vice president-IT
at HDFC Bank says, "The network storage design is outsourced to the hardware
vendor. We explain the current as well as future requirements to the vendor
and based on that the vendor suggests a storage architecture. The IS team takes
the final decision on the architecture."
The first step will be preparing a blueprint
of the enterprise based upon its current and future business requirements. Storage
vendors have professional services divisions that provide consulting services.
Avijit Basu, marketing manager, NSSO, HP
India says, "The first step will be to understand the company's business objectives
and goals and to conduct a storage assessment where we analyze the company's
present and future storage requirements."
While doing storage assessment, the consulting
team analyzes the kind of applications used by the customer, the number of servers
running diverse OSs and the kind of throughput required. The growth of data
and disk utilization has to be analyzed and mapped to the enterprise's business
goals in order to meet its future storage requirements. If the enterprise has
a distributed storage architecture, the first step will be to consolidate data
at single point for centralizing data management and simplifying backup procedures.
A vital point is to separate OLTP (On-Line Transaction Processing) data from
that generated by applications that aren't mission-critical. Qualification of
OLTP data helps in deploying business continuity solutions.
Mapping storage technology to the blueprint
Having identified the enterprise's storage
requirements, the next step is to select the appropriate storage technology.
For storing files in a departmental set-up where the applications aren't mission
critical, Network Attached Storage (NAS) is an ideal solution. These filers
communicate using Network File System (NFS) for Unix and Common Internet File
System (CIFS) for Windows environments. FTP, http, and other networking protocols
can also be used with NAS filers. NAS is platform independent, and using a filer
improves network performance as the NAS device is an appliance, and consequently
it performs better than a typical server would in the same role. NAS is typically
used for storage consolidation on a LAN.
If an enterprise is thinking of storage
for a back-end or OLTP application, that's where a SAN comes into the picture.
SANs are networks that are dedicated to data storage. Unlike in the case of
NAS, a SAN is separate from the LAN. Companies benefit by moving storage related
traffic off the LAN onto the fiber channel-based, SAN that offers faster performance
and lower latency. SANs also let companies opt for remote storage (up to 10
Due to their fundamentally different technologies
and purposes, it's not necessarily a choice between NAS and SAN. Either or both
can address an enterprise's storage needs. In fact, the lines between these
technologies are starting to blur. Down the road companies may choose to back
up its NAS devices to its SAN, or attach NAS devices directly to the SAN. This
is exactly what Orange does for its network storage infrastructure. Vinod Sadarangani,
vice president-IT, Orange says, "We are using NAS as an online secondary storage
and backup option instead of tapes."
Can NAS and SAN be used together?
DAS (Direct Attached Storage) is losing
ground to NAS and SAN. A combination of NAS and SAN termed as Enterprise Storage
Network (ESN) makes sense when an enterprise has different applications and
bandwidth requirements within its IT infrastructure. For instance, consider
an engineering company that manufactures a product. This hypothetical company's
messaging storage can be done on a NAS filer while the Supply Chain software
(deployed for the manufacturing set-up) can use a SAN. The same is true in case
of banks. A bank can rely on NAS for non-mission critical applications, and
SAN for its core banking application.
Ravishanker of Bank Muscat says, "ESN is
picked based on the application. We are planning to implement both NAS and SAN."
For mission critical applications such
as core banking, enterprises can implement a SAN and for storing e-mail and
hosting intranet data, they use NAS.
Sadarangani of Orange says, "Ultimately
NAS and SAN will merge. An enterprise SAN over Fiber Channel (FC) provides dedicated
point-to-point connections between servers and disks without any IP contention.
NAS isn't designed to take heavy loads, it would choke the LAN network."
Business continuity planning and DR
Enterprises need to identify their backup
requirements and define critical parameters such as recovery time objective
(RTO) and recovery point objective (RPO).
"Business Continuity Planning also includes
service level agreements for restoration of data. There should be flexibility
regarding what data is to be recovered in case of a disaster. OLTP and mission-critical
data should take the priority over e-mail," says Rawtani of EMC. Bandwidth availability
for recovering data is crucial. Trying to simultaneously recover all of an enterprise's
data can end up choking the fattest data pipe.
ROI on Storage
CIOs need a comprehensive plan to address
the high cost of managing distributed storage, or storage will remain a budget
buster. That's why they have their own way of justifying investment in networked
storage to top management.
Kameswar explains that being a semiconductor
design house, TI generates huge amounts of data that's shared and accessed by
its chip designers. "If I'm able to provide high-uptime and availability to
our chip designers when the load peaks, that's my ROI on storage. Storage has
to be as seamless as networking."
Though ROI in terms of money may not be
possible to define, there are tangible and intangible benefits that help justify
investments made into networked storage.
MSSP (Managed Storage Service Provider) as a concept has failed due to
security concerns. The fear that an MSSP might disclose customer or business
related data to a company's competitors is one reason why enterprises
have shied away from the MSSP concept. Also, the lack of consistent bandwidth
has been an impediment. MSSPs do not make sense in the Indian context.
According to Balasubramanian, "For a bank, data is criticalit
is our data on our customers. This is an important asset to us and we
need to maintain confidentiality. We own the storage ourselves but have
given only the support function to an outsourced agency with proper access
controls. We believe that MSSP is not a viable solution as of now."
Shailesh Agarwal, Country Manager-Storage of IBM India says, "Customers
wanting to outsource storage to MSSPs have doubts about the reliability
of remote connectivity."
Hosted storage services will not catch up in India as these models have
failed world-wide. However, there is an important trend seen in the Indian
storage market, that's managed DR services. Some enterprises are willing
to outsource their DR site allowing storage service providers and storage
consultants to manage their DR infrastructure. For instance Netkraft has
signed a deal with Movinture to design, maintain, and manage its DR site.
Netkraft's DR site will be hosted at Movinture's own data center and they
will manage the entire storage infrastructure for Netkraft. Similarly
Tata Teleservices Ltd (TTSL) has invested in EMC's storage infrastructure
and it offers managed DR services to Indian companies. Asian Paints is
using TTSL's storage infrastructure for DR.
When it comes to backing up data there are areas where an enterprise
prefers to hire the services of storage integrator, especially for managing
Service Level Agreements (SLAs), and for managing tapes. DAS is still
predominant in India and backup is often taken manually. A typical manual
system consists of a single tape drive connected directly to a server
for backup. A network administrator loads and unloads tapes physically
while executing backup procedures. Sometimes people change, load or swap
the wrong tapes. In such a system, it takes a long time for restoration
and searching through stacks of tape cartridges to find the right tape.
This long backup window prevents organizations from carrying out other
activities until the backup is complete.
Texas Instruments India has invested in Quantum ATL libraries and this
move has aided the company in reducing its backup window from 72 hours
to 9 hours or less. Kameswar says, "We are able to backup our entire
data in less than 9 hours. The ATL library offers faster backup and retrieval
of data and it has helped in media management." Earlier TI was spending
half a million dollars on media management.
Balasubramanian of HDFC bank adds, "We have a tape library containing
LTOs with which we are able to backup from multiple servers (that are
attached to SAN). Labeling and cataloguing is easier. We are able to back
up multiple systems onto a single LTO library. This media is more reliable
than DAT or DLT. We are able to operate at higher speeds with greater
reliability and at the same time reduce the overall spending on media."
Handling and off-site storage becomes easier as the number of tapes used
- Know your business
Business Goals, Corporate competency, business uptime, RPO, RTO
- Analyze what you do
Analysis & break-up of data management tasks
- Know the exact process
Full time equivalents, impact analysis, response time
- Know how much you spend
Cost of ownership
- Reduce footprint (use fewer boxes)
- Buy a lower number of software licenses for replication, backup and
- Reduce maintenance contracts (including costs of contracts)
- Reduce staffing levels
- Reduce network/SAN infrastructure
- Reduce total cost of ownership (TCO)
| NAS (IP based)
|| SAN (Fiber channel
|| CAS (Content Addressed
|| For file sharing
and a low-cost option for hosting databases.
|| For transferring
block data between servers and storage devices
| For fast access to
read only information
|| Economic advantage
|| Lowest cost to connect
and share files
| Takes the load of
moving storage data off your LAN letting you perform server-less backups.
|| Lowest cost online
|| Typical applications
| Software development
houses, product design companies, for hosting small databases and data generated
by workgroup applications.
|| OLTP, data warehousing,
ERP, large databases.
|| E-mail archiving,
content management, document retrieval
| Source: EMC
Akhtar Pasha can be reached at Akhtar@expresscomputeronline.com