In 2010, Indian auto player Mahindra Vehicle Manufacturing limited (MVML) faced a tough call—spend Rs 2 crore in 40 servers, or invest Rs 50 lakh in server virtualization architecture. In addition, MVML’s Greenfield facility (built in 2007) based at Chakan in Maharashtra had obstacles like decentralized infrastructure to overcome. This is about the route MVML chose to deploy virtualized architecture, and its consequences for the facility with an annual capacity of 3.2 million vehicles.
The Chakan Greenfield plant’s first phase is part of MVML’s development plans for a facility that will eventually encompass 700 acres. The company started at a grassroots level with 280 acres of a land, and no water, electricity or infrastructure. “We had to get IT infrastructure right at the first go, and create a benchmark for all the other phases,” says B Venkatakrishnan, the head of IT at MVML, Chakan. There were 11 shops with 11 different server rooms and several applications. This decentralized infrastructure made it tough to provide round the clock IT services.
Virtual servers to the rescue
In 2010, MVML’s plant had to integrate applications provided by Rockwell Automation Services, and required 40 additional servers. At this point of time, MVML found itself battling improper connectivity. “We faced fiber cuts, communication issues and GSM service failures quite often,” says Venkatakrishnan. MVML identified server virtualization architecture as the antidote for decentralization.
- vSphere Enterprise plus 4.0 – 24 CPUs
- vCenter Server 4 Standard – 2
- vCenter Operations Standard 1.0 – 50 CALs
With a capacity of 3.2 million vehicles per year, stakes were high. A single minute’s downtime in server virtualization architecture could incur huge losses. MVML took buy-in for server virtualization from key stake holders such as the management, department heads and system integrators (like Rockwell).
MVML evaluated server virtualization solutions from VMware and Citrix. Industry trends, feedback and Gartner reports helped MVML narrow down on VMware vSphere ESX and vCenter management servers.
Platform 1: Four HP Proliant DL380 G7 (vSphere ESX Server) servers
DL 380 rack server with two Quad Core X5550, 64 GB Memory, 4* 300 GB HDD, 2 NICs of 4 ports each, DVD, four RPS units
Platform 2: Two HP Proliant DL 580 G7 (vSphere ESX Server) servers
DL 580 rack server with four X7542 CPU, 96 GB of Memory, 7*300GB HDD, additional eight port NIC, three BBWC along with RPS
Platform 3: One HP Proliant DL580 G7 (vSphere ESX Server) server
DL 580 rack server with four Quad Core E7520, 128 GB Memory, 4* 300 GB HDD, two NICs of 4 ports each, DVD, RPS
Platform 4: One HP Proliant DL380 G6 (vCenter Management Server) server
DL 380 rack server with two Quad Core X5550, 16 GB Memory, 2*146 GB HDD, 6*300 GB HDD, 4 port NIC, DVD
Moreover, 99% of MVML’s applications were VMware compatible. Good administration support from the backend and vMotion monitoring features further boosted appeal.
MVML’s server virtualization project started in December 2010. A two month proof of concept (POC) with 20 virtual servers in the test and development environment came first on the roadmap. MVML created 600 scenarios to eliminate potential risks, and completed the POC in March 2011. MVML’s server virtualization architecture went into production in April 2011.
As part of phase one, MVML deployed eight physical servers hosting 40 virtual machines (VM). These VMs host all applications and the database. VMware and Rockwell Automation helped MVML’s internal team during the implementation.
In phase two, MVML will deploy three physical servers on the VMware platform to host 70 VMs for all new services, as well as non mission critical and legacy applications (like the training center).
To battle connectivity issues, MVML deployed a three-tier network architecture based on High-Level design using Cisco platforms. According to Venkatakrishnan, the three-layer design provides high availability using redundant hardware, software and network connections. It incorporates automatic procedures for recovery and reconfiguring of network paths during failures.
The three core networking functions of this architecture are:
- Network security based on the Cisco Self-Defending Network
- IP-based communications
- Mobility and wireless LAN services
Virtualization of network architecture also enabled multiple distributed virtual networks with three logical functional areas:
- Access control
- Path isolation
- Service Edge
Platform: Hitachi AMS 2100 Modular
Hardware configuration: 64*300 GB SAS 15K RPM HDD, modular 4 GB Cache, 2*1 GBPS iSCSI I/F Board / Power Cable 250 VAC 10A IEC320-C14 / basic operating system - modular / AMS2100 Controller with two embedded 8 Gbps FC port / SAS/SATA storage expansion tray
Logical configuration: Hitachi Dynamic Provisioning Base Kit / Hitachi Dynamic Provisioning License / In-System Replication Bundle Base Kit / In-System Replication Bundle License
Money well spent
According to Venkatakrishnan, MVML’s server virtualization architecture has reduced administrational overheads, as the company requires just one administrator (as opposed to six earlier). Provisioning of servers takes minutes instead of weeks. Server maintenance doesn’t interfere with application performance, as VMs can be moved seamlessly from one physical host to another. Server virtualization based architechture made it possible to avoid cooling investments in additional precision airconditioning units. MVML has also deployed a building management system application to enable monitoring of power, cooling, and carbon footprint.
According to Venkatakrishnan, new technology is always accompanied with ambiguity. “Gearing our workforce for virtualization was difficult. We sent them for training and industry events to gain insights on virtualization, as well as be prepared to manage production environments.” During the first year, MVML did not face server virtualization architecture related issues.
Management tools like Nagios for availability of data, WhatsUp Gold for networking and vCenter operations have been deployed by MVML. If a service goes down, an SMS alert is triggered to the first level support group. Depending on severity and availability, SMS alerts are triggered to the various agencies as well.
The second phase of MVML’s Greenfield project, along with the virtualization project’s second phase is set to begin soon. MVML also plans to host all legacy applications like antivirus, LAN monitoring servers and file servers onto the virtualized platform.