Server Update 2006

Tales from the data centre

Whether it is server management, seamless scale-ups or DR, there is always the need for role models to look up to. Here we present three organisations which share their server implementation success stories

When it comes to server-front victories, this time’s server update decided to look at the biggest of them all. With this objective in mind we narrowed down the field to three organisations.

Hindalco, the metal segment leader, comes first on the list. The organisation’s rapid growth meant that the ability to scale up as fast as possible on the server front is the biggest challenge. The route taken by Hindalco to achieve this is a case study worth looking at.

On the other hand, Ashok Leyland faced the server consolidation challenge. This is a story of how the organisation improved itself in terms of availability as well as productivity on the process and end-user fronts.

Indian Oil comes next with its two-tier DR implementation. Their’s is a story of how to use a categorised approach to DR.

The scalability factor

When it comes to catering to the IT requirements of an expanding organisation, better, faster and more scalable servers are a must. Rishiraj Verma reports on the server implementation at Hindalco

Sanjeev Goel

When an organisation is the size of the metal manufacturing major Hindalco, technology is crucial. This, according to Sanjeev Goel, the company’s Senior Vice-president for IT, is the major reason for the implementation of the newer servers.

“If we’re talking about the quantity of performance, we currently have around 2,500 users and we are planning to move up to 10,000 by 2010,” reveals Goel. He explains that the number of business applications needed to continue operations were constantly on the rise.

The scalability hurdle

Goel states that application integration became essential as it was getting increasingly difficult to manage and deploy applications due to the organisation’s size. “Achieving scalability was one of the biggest challenges that we faced.” Thus arose the need for scalable servers. The company put together a robust disaster recovery management plan that called for the use of high-end servers.

The best option

“We wanted to look at servers that would be easily scalable,” is Goel’s reply when asked about what kind of solutions the organisation was looking at. IBM’s Power 5 series servers seemed the best option for Goel and his team at Hindalco. “This technology helped us get into rack mode.” He explains by saying that more server modules can be added as per need. These servers can scale vertically, which saves space as well. Hindalco has chosen p570 DB servers with Oracle 9i configured at the data centre, and a DS 8100 storage array with a capacity of 4.5 TB. The OS is IBM AIX 5.3.

Goel says that IBM was chosen since the “value for money” they provided with the Power 5 architecture was greater than that offered by the competition. He says that IBM has a clear roadmap starting with its P3 range of server processors, moving on to P4, now P5, and the P6 which is expected to come up in the near future.

He says that a DS 8100 has been installed at the secondary site to ensure fast recovery in a disaster situation. The servers have features to make sure that problems at the chip or memory level can be handled by the servers itself. “These multi-application servers take care of our increasing application deployment needs.”

The timeline

According to Goel, the project was phased, and it was made sure that every phase was given its due importance. “The project started in September 2005 with a branch integration project. After evaluation and synchronal DR, the servers and SAN were in place by October.”

This was the deployment’s more important phase when major infrastructure placements and data movement took place. “Between October 2005 and February 2006 we have been able to add 32 new sites,” he says as he informs that the entire project was completed within five months.

Several benefits

“Availability has gone up,” is the first response from Goel as he lists the benefits Hindalco has been able to achieve because of the implementation. He says that the implementation has resulted in zero downtime and no data loss whatsoever. Plus it has gained access to easily scalable compute power.

Pre-implementation, flash copies of data had to be made. “We can now take multiple images for the clone and test environment.” Goel also states that for an international organisation like theirs, online visibility is very important and this implementation has increased it.

When asked about the business benefits, Goel says that the company is currently in the closing stages of project stabilisation. According to him, this is the reason that the project’s savings haven’t been quantified yet. The financial benefits should be out in three to six months.

On the horizon

Goel says that for now the architecture has been finalised and that the organisation will stick to it unless there is need for upgradation. “This was the reason the rack model was chosen. More modules can simply be added without the need to actually upgrade the servers themselves.”

As far as the future is concerned, he says that the company is planning to touch the five billion dollar turnover mark, but that this should not call for a technology upgrade very soon.

Consolidating for productivity

Ashok Leyland wanted to enhance productivity and improve the dissemination of IT and business process knowledge among employees. Server migration and consolidation helped it achieve its goals. Dominic K reports.

N Mohanakrishnan

As part of an enterprise-wide initiative, Ashok Leyland went through internal process restructuring in the late 90s. This restructuring took place in areas such as supply chain, forecasting and outbound logistics. N Mohanakrishnan, Special Director, Information Management Services, Ashok Leyland felt that the only way to get the best out of this huge restructuring exercise was to migrate all of the company’s processes to a common IT platform.

“We were clear in our objectives to have the best-in-class information technology infrastructure and services for our growing businesses,” explains Mohanakrishnan.

Network failure = no ERP updates

Prior to setting up of the centralised data centre at Ennore in Chennai, Ashok Leyland had a rudimentary ERP with systems operating from individual locations and a data consolidation facility on a periodic basis. There were several issues that needed to be sorted out because of this arrangement.

The main problem was that the ERP application did not have a provision for manual updates in case of network failure. This clearly meant that the business activity on hand would have to be deferred until links were restored.

The integrated route

All of Ashok Leyland’s business processes now run on a single integrated ERP solution. This solution was developed in-house between the years 2000 and 2003 along with the help of HTMT.

The WAN has 110 points of presence across the country and all these points access the ERP system hosted at the data centre in Ennore for daily operations. The entire infrastructure including the WAN is redundant to ensure that there is no single point of failure facility at any point of time.

The company uses Microsoft products such as MS Exchange, MS Office, Sharepoint portal server and enterprise project management solutions. A centralised help desk has also been set up at Chennai to support 75 locations that support desktop management for 3,000 desktops and laptops.

Hooking it all together

All major locations of the company such as manufacturing centres, and corporate and marketing offices are connected by a 2 Mbps leased line with redundant ISDN connectivity from more than one telecom service provider.

Other points of presence are connected using 64 to 256 Kbps leased lines depending upon the level of activity and load. Alternate connectivity is provided through ISDN dial-up with appropriate bandwidth and the smaller locations use Very Small Aperture Terminals (VSAT) connectivity.

The new ERP solution, developed in-house, has been designed such that all the dealer points and the service outlets meeting customer needs will be able to access it. On completion of the total initiative, Ashok Leyland’s WAN and dealer chain will encompass 400 points of presence.

Consequently, the IT platform’s continuity for carrying out business processes became critical. Steps were taken by the company to commission a business continuity data centre in one of the major plants at Hosur. This will be operational in the course of next couple of months.

Once the upgradation at Chennai and the Hosur site are commissioned, the company will migrate its solutions to high-end HP storage platforms.

Information security

In terms of security, the company has designed and deployed an infrastructure which has virus protection, firewall and intrusion detection. All of them are running on various HP servers. The entire Web-based services of the company are Oracle-based solutions and run on Compaq Alpha hardware.

The company has put in place information security policies and procedures. It is also certified for BS 7799 by Standardisation Testing and Quality Certification, a division of Department of Information Technology, for compliance with BS 7799 guidelines.

In addition to this a number of security tools and intrusion detection and prevention solutions protect the company’s data. Continuous monitoring and evaluation of adequacy of the information security is made and necessary augmentation/modification to security infrastructure/policies are put in place from time to time.

Additional initiatives

Besides covering the entire organisation on a common IT platform, the company is working on two major initiatives. One is to cover dealers and customers and the other one to manage the supply chain.

These are enabled through customer-and dealer-related processes and supply chain processes, which are migrating to concurrent engineering in phases. The solutions were developed in-house.

Other initiatives include deployment of hardware for enabling the company’s messaging, office automation and enterprise project management, and portal services besides virtual campus and Six Sigma.

These are presently deployed on Intel Xeon-based HP servers and are in the process of being upgraded and migrated to blade servers. Server Management supports more than hundred servers of HP Integrity and HP Unix.

HP has provided the end-to-end IT infrastructure service and management that include 24x7 support services. HP has also provided multi-vendor support with project management to 70-plus locations of Ashok Leyland bound by Service Level Agreements (SLAs).

Benefits and ROI

The deployment has enhanced process productivity with higher infrastructure availability to business. It has also improved end-user productivity.

Elaborates Mohanakrishnan, “The impact of IT on business and vendor partnership in this endeavour has been tremendous. The cycle time has come down drastically, and process productivity has gone up by about 30 percent. Manufacturing planning is also conducted automatically now.” There is higher infrastructure availability with hardly any downtime. “The greater predictability of spending on IT and IT assets means that our TCO is better than before,” he adds.

The benefits derived with IT as the centre point is being felt across Ashok Leyland. The knowledge of IT and its impact on processes and operations has improved across organisational hierarchies. Communications across the 100-plus locations has become comparatively faster in the changed environment.

Indian Oil's two-tier advantage

Indian Oil needed high-end servers for running its enterprise-wide applications that were secured against disasters. A two-tier architecture helped it achieve its goals. Dominic K reports.

S Ramaswamy

In 2002 Indian Oil Corporation Limited (IOCL) realised that deploying solutions on a centralised architecture has advantages and disadvantages—technical as well as business. Any failure of the centralised database server would bring IOCL’s business to its knees. This was the point when IOCL decided to provide sufficient availability of SAP R/3 services to its operators. The solution was to design and deploy a two-tier disaster recovery (DR) architecture.

As S Ramaswamy, GM, IS, IOCL informs, “Data storage and back-ups on servers form a crucial part of our mission-critical operations that need to be addressed by an advanced architecture and related support services.”

One of the world’s largest

IOCL’s SAP R/3 installation is one of the largest in the world. It hosts over 5,000 simultaneous users from 600 locations. The systems and services operate on a centralised architecture and rely on a centralised database system. This helps IOCL users across the country to connect to centralised database servers.

Today, the deployed architecture’s sole objective is to maintain data integrity, DR and critical business management. “The deployment complements business continuity (BC) at IOCL with minimal downtime. Data losses are eliminated or minimised at the very least,” says Ramaswamy.

For every eventuality

IOCL’s DR plans are well managed. Various disruptions and disasters have been systematically categorised and documented at IOCL. Disasters are categorised as local failures, logical data failures and site failures. Local failures include CPU faults, disc and array failure, failures in LAN and SAN, and software and utility failures. Logical data failures include system crashes due to software bugs, data corruption, accidental deletion and application of wrong patches. Site failures include power and telecom outages, acts of war or terrorist attacks, and human errors.

Based on disaster criticality, the SAP R/3 services are provided from four different centres. Under no-disaster conditions, the services operate out of IOCL’s Corporate Data Centre (CDC) for the primary production system. The CDC is located at the Indian Oil Institute of Petroleum, Gurgaon.

The R/3 production environment constitutes a database server with an R/3 central instance server and 15 R/3 application servers. The deployment also includes various enterprise-class storage sub-systems and a tape library over a high-speed LAN and SAN.

The database server and the R/3 central instance server run on HP Superdome 9000-class machines with 18 HP PA-RISC 8700 series CPUs. 24 GB of the server’s main memory is connected to the HP XP1024 storage sub-system and HP ESL tape library over the SAN through an HP FC-64 SAN switch. The database server currently runs on Oracle Database Server Version 9.2.0.5.

Individually, every SAP user is connected to the production system. This is achieved through any of the 15 SAP R/3 application servers. These are deployed on 13 HP Series 5405 servers (3 CPUs, 6 GB RAM) and two HP series 7410 servers (8 CPUs and 24 GB RAM).

The Corporate Communication Centre acts as the primary communication hub under normal conditions. The centre is also located at Gurgaon, but in a separate building at a cable distance of 300 metres from CDC. A baseband and radio frequency (RF) solution is also provided within the same designed architecture for redundancy. This centre forms the backbone for various e-commerce applications, SAP, FTP and e-mail.

Finally, there is the Metro Disaster Recovery Centre (DR1) to be used as a production system during a localised disaster at the CDC. It is located within the communication building at Gurgaon. The DR1 centre has an HP 9000 Superdome server with 12 CPUs and 16 GB RAM. This centre also hosts five HP L-class servers (three CPUs and 6 GB RAM) running as application servers. It is configured for the primary production environment. IOCL’s development and quality server hosted at DR1 is on an HP L-class server (three CPUs and 6 GB RAM). Cisco’s network node manager and HP OpenView software manage the communication centre.

Twice the strength

The two-tier recovery architecture aids faster and consistent recovery. It is also expected to support communication and SAP R/3 services during natural and man-made disasters.

The DR1 operates from the communication centre within the complex, and lies about 300 metres from the production data centre.

The Main Disaster Recovery Centre is based at the Sanganer terminal in Jaipur. It is located about 250 km from the Gurgaon facility, and is known as the Business Continuity Centre. This centre has the responsibility to maintain BC in case SAP R/3 services stop. The primary site at Gurgaon, which also hosts the data centre, became operational in November 2002. The secondary DR site at Jaipur has been operational from early 2005.

Inside IOCL's network

The network topology is based on hub-and-spoke. This also connects the four divisional and regional headquarters, refineries and state offices directly to the communication centre at Gurgaon. Connectivity is established by means of high-speed PAMA (Pre Assigned Multiple Access) VSAT links and 2 Mbps leased lines. 402 leased lines make up the overall network.

Data security and communication in the network and between various servers is managed via Cisco 6509 Layer-3 switches. These support and maintain the data centre network backbone. A 1000 Mbps network on fibre cables takes care of the network traffic between the application servers and the database server, while a 100 Mbps network maintains traffic between the end-users and application servers. The LAN is connected on a single Cisco 4507 Switch.

A clustering tool called MC Service Guard is deployed to ensure availability and continuous running of database instances and applications. The storage between tiers is networked via two HP 2/12 StorageWorks SAN Switches. The primary and the secondary sites store about 40 TB.

VSATs to the rescue

VSATs form the primary communication medium while ISDN and PSTN lines form the back-up connectivity. About 326 VSATs help users communicate through the central primary communication hub. The corporate-wide VSAT network is on INSAT 3B using the Ku band. The organisation has hired one full transponder for the same.

All divisional offices, terminals, depots and LPG plants are connected to their respective state offices with 64 Kbps (or n x 64 kbps) leased-line links. These are also connected to the primary communication hub at Gurgaon using the TDMA and FTDMA network. The data redundancy on multiple servers based on the two-tier architecture has enabled Indian Oil to maintain its zero downtime expectations with better productivity and an always-available system.