Project Log

Redesigning HDFC Bank’s SAN

Maintaining a multi-vendor storage area network (SAN) was becoming difficult at HDFC Bank, following which it redesigned and consolidated its storage network. Harish Shetty, VP for HDFC Bank tells Megha Banduni how the bank's SAN was redesigned.

"The performance of the IT systems have accelerated by 30 percent after this deployment" - Harish Shetty Vice President HDFC Bank

HDFC Bank has a network of 531 branches in 228 cities across India, as well as 1,500 ATM sites, all of which are linked in real-time to accelerate the flow of data and transactions.

The bank has to manage a large amount of customer transactions and ensure that data is stored for no less than seven years.

To satisfy this need, we built a multi-vendor storage area network (SAN) at our headquarters with storage solutions from Sun Microsystems, Hitachi Data Systems, HP and IBM. The devices are connected to edge switches and are linked to a core consisting of early-generation Brocade Silkworm Director switches. The core, in turn, is linked to a 300-slot tape library in Mumbai. The back-up solution consists of Brocade edge switches and LTO generation 1 and 2 tape solutions. We access information either from the storage subsystem or from tape, depending upon the source application.

At HDFC Bank, storage is divided into five categories—mission critical (24x7), critical but not 24x7, large amount of data (non-critical), testing and other non-critical data as well as archival and back-up.

All the bank’s transactions are online. Hence storage plays a critical role in our business’ performance.

For back-up, we use a tape library. Earlier, it took an hour to back-up, now it takes 10 to 25 minutes. As per regulations, we keep data on tape for eight years.

We rely upon a Sybase database running on Solaris for servicing retail customers and Oracle databases running on AIX, as well as a variety of other applications designed to support and sell various financial products.

Managing data smartly

With millions of transactions taking place daily, such as loans, deposits, and checks being processed, the bank wanted a system that could handle terabytes of data and scale up. Also, issues related to the maintainence of SAN infrastructure, operations and costs, were other key challenges.

As per regulations, data has to be stored for at least seven years. Managing huge data stores for such a long period was another hurdle.

Also, with the growing volume of operations and processes, our storage requirements were growing day-by-day.

Whenever divisions within the bank demanded additional storage, we added servers and edge switches to the SAN.

As a result the storage infrastructure was becoming difficult to handle. Before reaching its destination (server and tape library), data often had to pass through many devices within the SAN resulting in the creation of bottlenecks.

As a result, we decided to redesign the SAN. The number of inter-switch links (ISLs) had increased exponentially and it became necessary to reduce complexity within the SAN. This was an operational issue because of support and management issues.

We always had a fear that our SAN could bring down the entire banking process. To overcome this, we streamlined our storage infrastructure to bolster its reliability as well as its performance. Moreover, it became important to increase our existing 752-port capacity to meet our rising storage needs. Our continued success in the financial markets depended on finding the right vendor with the right solutions, service levels and expertise.

Speed and simplicity

We examined solutions from Cisco, McData and Brocade and finally chose Brocade whose solutions offered the best migration path, allowing us to improve the overall performance of our storage for data warehousing while continuing to leverage our legacy systems. Moreover, the entire redesign was smooth and non-disruptive to our users. We currently rely exclusively on Brocade switches for all of our storage requirements.

Brocade examined how our business works and it sought to meet our needs by looking strategically at the big picture and proposed a new storage architecture based on the industry’s leading core switches, complete with substantiated technology proof points.

With support from Brocade’s engineers, we redesigned our storage infrastructure to streamline data traffic. We deployed six Brocade Silkworm 48000 Director solutions, all linked to a single fabric. Each of the switches offers 256 ports, allowing us to support thousands of storage devices.

The 48000 Directors deliver four gigabit per second, double the bandwidth of our legacy switches. This helps mitigate latency due to increasing traffic, as well as take full advantage of our high-density storage servers. We also eliminated many ISLs by connecting vital application servers directly to the core.

Optimising storage performance

We use Brocade Advanced Zoning that is supported by the Brocade Fabric OS- the operating system for the core switches, to localise traffic on the fabric. Each application server is linked to its own storage solutions by no more than one SAN switch, thereby containing traffic on the fabric to individual islands.

For example, servers that generate data intended for the tape library are linked to an edge switch dedicated to support the library. Similarly, departments that often need to retrieve data from the tape library are linked to that system’s edge switch, providing direct and speedy access to information. As a result, traffic bypasses the core, eliminating many hops, conserving bandwidth throughout the fabric, thereby smoothening data flows.

Later, we installed ten Brocade SilkWorm 4100 Fibre Channel switches to connect application servers and storage devices. These high-availability devices support 1, 2, and 4 Gbps speeds, ensuring compatibility with the legacy switches. Additionally, they provide ports on demand, allowing administrators to scale the platform’s 16 ports to 24 and then 32 with license activation.

Today, all key banking applications at HDFC bank have 4 Gbps links to the SAN, while systems that aren’t critical get 1 or 2 Gbps connections optimising storage performance based on business needs.

Additionally, we are able to reduce the number of hops for our most important data flows by consolidating departments and applications into localised SAN islands.

Vendor support

Brocade provided technical training and other services and assists our IT staff. For example, administrators use Brocade WEB TOOLS, an intuitive, browser-based application to centrally monitor and manage the Brocade switches, and Fabric Watch, which enables each switch to constantly watch its SAN fabric for potential faults and automatically alerts network managers to problems before they become failures.

In addition to this, they can centrally configure the entire fabric, specific switches, and even individual ports to ensure that the infrastructure performs effectively.

In reality, we don’t spend much time watching our Brocade switches because they function reliably. Their rock-solid performance lowers our administrative burden and reduces the cost of management.

30 percent performance boost

Storage performance has gone up by 30 percent after deploying Brocade’s solutions. Moreover, with 1,744 ports supporting over 200 terabytes of storage capacity, the SAN is scalable enough.

Our storage has grown almost 300 percent from 2000. We have benefited in terms of garnering more business, maintaining response time and reducing downtime. System availability has improved dramatically and MIS, reporting, and the analysis has become more comprehensive.

Brocade has provided us with a vastly improved architecture and 4Gbps technology that has given us good performance, reliability, and investment protection that a financial institution of our size requires.

The return on investment on account of redesigning the SAN is in terms of enhanced productivity and lower operating costs.

Today, we effortlessly archive three terabytes of data every day. Users can access stored data rapidly because of the SAN’s efficient design, ensuring that they always provide customers with the finest service.

Moreover, the entire storage infrastructure is cohesive and dependable eliminating technical issues that threatens to reduce system availability and adversely affect profitability by hampering banking operations.