with the problem of managing exploding data storage
requirements, Texas Instruments India chose to implement
a combination of DAS and NAS. by Prashant L. Rao
Instruments (TI) was the first MNC to set up R&D
operations in India's Silicon Valley. Over the years,
the company's Bangalore setup has grown to become a
major design center. Along with its growing success
in creating chips for a global audience, TI India's
data requirements have grown apace.
N.V. Kameswar, General Manager, Information Technology,
Texas Instruments India says, "The way storage
has grown is commensurate with the growth of design
projects at TI India. Not only has the number of projects
gone up, the complexity of projects has increased significantly.
People who used to ask for gigabytes are now asking
In December 2000, TI had 3 TB of storage. By the end
of 2001 that had gone up to 9 TB, and by the end of
this year it is expected to be 17 TB. Personal data
in the form of e-mail, home directory files etc account
for less than 1 TB. The bulk of TI's data consists of
design project files.
TI India uses a mix of Solaris and Linux servers; the
server pool consists of 900 CPUs. The server protocol
used is NFS and NIS is used for providing IPs and to
reference data areas. As a result, TI looked for CIFS
and NFS support. Another critical requirement was the
ability to take snapshots.
Texas Instruments India was the first MNC to
set up an R&D base in the country. It has
a sprawling development center in Bangalore
that designs chips which go into end-user equipment.
The success of TI India's R&D efforts lead
to a data explosion. Its data was growing exponentially
(triple in 2001 and close to double in 2002).
The company needed a storage mechanism that
was more scalable and robust than DAS and which
would work with its server pool of Solaris and
As most of its data was file data, choosing
NAS boxes was logical. TI India went in for
NetApp filers on recommendation from its global
Today, TI can scale its storage pool with ease,
storage is centralised, redundancy exists and
backup processes have improved.
As a design center with high-end computing requirements,
TI India needed to store its compute farm data. NAS
was a logical choice as most of TI's data is file data.
"A SAN will not work because we don't read data
off a disk array, it's all Unix file data," says
Tony Idiculla, Lead Engineer, DCE & IT, Texas Instruments
next step was to choose a vendor and TI chose NetApp.
"The global strategy team recommended NetApp filers
as part of global standardisation. The aim was to have
a common operating environment and avoid reinventing
the wheel," says Kameswar. "People design
chips and we help them. To make that happen we need
Standardisation helps transfer data between TI's many
TI India hosts its data in a mix of NAS (NFS+CIFS) and
DAS (FC-AL based) devices. However, its predominant
platform is NAS, hosted out of Network Appliance filers.
The company did a pilot that lasted between two and
three months. F760 filers were used for the evaluation
(an old generation box from NetApp). The pilot deployment
consisted of 3 TB across two of these filers.
The actual deployment took ten days. Six F840s were
installed at TI. Today, TI Bangalore has deployed six
F840s and four F880s. The first rollout was done by
teams from TI and NetApp. Wipro Infotech was selected
to handle the AMC and facilities management. The company
has purchased 47 TB of raw capacity across 10 filers
in five clusters. Of this, 35 TB is the usable space
with the remainder being used for creating file systems,
creating a RAID 4 system. TI is using a RAID group size
of eight, which is the standard recommended by Network
This implementation has given TI India a common storage
pool where any server can access data stored on any
filer. That said, the bulk of TI's data is project-based
and stored in specific locations. There is an 'application
pool' that most of the chip designers access. This application
pool is handled by a filer or a set of filers. In each
filer the data is pumped through the head. Disk arrays
are attached to the NAS head. There are two filers in
a cluster supplying data using NFS (Solaris) or CIFS
(Windows). If one NAS head fails the other takes over,
providing data from both filers. The redundancy exists
only from the processor perspective, data is still readable
if a processor fails. This failover works down to the
Ethernet MAC address.
"You can scale up by adding disks, this is the
advantage of using RAID 4," says Idiculla. Today
TI uses a mix of F840s and F880s. The F840 was introduced
in December 2000. The F880s were launched in 2001 and
TI procured them as well. The F880s have superior capacity
and better processors in addition to which they support
dual processors. The 880s also have newer firmware and
scalability is higher in a cluster environment. The
F880 offers a maximum raw capacity of 16 TB. NetApp's
Data ONTAP software can scale up to this level.
After implementing NetApp's storage solution, TI appreciates
the following benefits:
Interoperability: NetApp's filers work with
PCs and Unix servers.
Snapshots: Using this feature, user home directories
are backed up each night (recycled).
NDMP (Network Data Management Protocol) 'serverless'
backup: Using Veritas backup software taking advantage
of NDMP lets TI backup without affecting server load.
Block level transfer from filer to tape: The
NetApp filer can transfer data in block mode to tape
Scale up on the fly: Both disks and shelves
of disks can be hot-plugged. The older FC9 racks have
seven drives and two power supplies. The newer racks
support 14 discs and two power supplies.
Looking forward, TI India would like to focus on scalability
and disaster recovery (DR). "We are currently adding
approximately 1 Terabyte of storage every month,"
says Kameswar. So TI will have to keep adding disks,
racks and filers as demand increases. Over and above
that, there's the issue of disaster recovery. Today
there is no disaster recovery site. However, TI does
have offsite storage of backup tapes. A DR solution
is a high priority activity for TI India this year.
Prashant L. Rao can be reached at firstname.lastname@example.org