Circuiting the network needs

Font Size

Print

Feedback

Email

Discuss

: Growth pangs literally forced electronic design and automation major Cadence to reboot its IT infrastructure in India. Computing capacity demands were growing fast and the need for real time data transfer was also escalating.

Cadence has a centralised server farm compute infrastructure. It enables the company to run many regressions on its products before releasing them to customers. The products include software and hardware, methodologies, and services to design and verify advanced semiconductors, printed circuit boards and systems used in consumer electronics, networking and telecommunications equipment, and computer peripherals. Earlier, the entire compute server infrastructure was scattered across 12 small computer rooms in three buildings in its Noida campus. The tremendous influx in traffic due to bandwidth intensive applications forced Cadence to consolidate its 12 small-size server rooms into one large datacentre. The company wanted a smooth transition from the distributed network to a single datacentre to ensure its network was available to all its developers and end users.

The US-headquartered company, which provides electronic design solutions for advanced IC and systems design, wanted a sturdy network that would seamlessly link its sales offices, design centres, and research facilities spread across the globe and also scale up to meet its ambitious expansion plans. Availability, scalability and performance were the main criteria while devising the company’s IT network requirements. Cadence wanted an infrastructure that could be upgraded with ease and kept pace with its growth.

Encouraged by its trackrecord at the datacentre located at Cadence’s corporate headquarters in San Jose, the company opted for Foundry’s BigIron RX Series of backbone Layer 2/3 switches to handle the traffic in bandwidth. BigIronRX has the highest non-blocking architecture optimised for maximum throughput and low latency for all packet sizes. “Foundry understood our requirement of high-end grid-based network IO intensive infrastructure and positioned the right products which could meet our exact requirements,” says Ashwin Rao, IT group director, Cadence India.

The BigIron RX backbone switch aggregated over 500 of their servers with ample room for growth. This helped strengthen the datacentre’s backbone. Furthermore, each switch utilised non-blocking 48-port 10/100/1000 MRJ-21 Ethernet modules. This module supported high-density configurations that safeguard Cadence against compromising on quality of service, latency and throughput. The 48-port 10/100/1000 MRJ21 module also increased the non-blocking Gigabit Ethernet (GbE) capacity to 768 ports for a single BigIron RX system and 2,304 ports in a standard 7-foot equipment rack.

Working alongside D-Link, Cadence faced no network congestion or downtime during the consolidation process. The datacentre can now house approximately 2,500 servers. The switching infrastructure acts as the core of the entire campus LAN . It has created enough headroom in the network to accommodate the expansion plans of the company. The network switch solution provides Cadence the required amount of endnode connectivity and uplink connectivity (in terms of density and speed) within an optimum number of chassis.

Foundry installed high-density configurations at the centre that suit high capacity server farms and high performance computing environments. The number of servers and compute nodes are driving demand for non-blocking interconnect solutions that can scale beyond 500 GbE ports per system. The company felt rising per rack power consumption due to technological advancements/trends would lead to more complex cooling and electrical requirements. So Cadence has set aside some space for this purpose.

The company plans to add more edge switches to accommodate additional servers due to increased demand for compute capacity. The company also continues to explore the possibility of deploying various server consolidation and virtualisation technologies to further enhance the compute capacity in the same physical space.

The company has already earmarked a space for the next phase of the datacentre that would be able to take care of much higher KW of electrical load per rack without compromising on reliability and uptime.

This would be crucial as blade servers that consume more power per rack and hence require more complex cooling infrastructure, and other similar technologies gain wider acceptance.

The Noida centre is connected to the global Cadence data network using DS3 links to meet real-time data transfer requirements. This ensures round-the-clock product development and R&D activities in its centres across the world. Cadence India uses a centralised storage infrastructure consisting of IBM ESS (Shark) and NetApp Filers. Based on performance and criticality of applications, a multi-tiered storage infrastructure is used. Client systems use a mix of SAN and NAS technologies to access centralised data. Leading tape technologies and backup management softwares are deployed to ensure no crucial information is lost. “Being a product-oriented software company, security of our IP and our customer’s IP entrusted upon us is the top most priority. CSI (Cadence Security Initiative) is one of the key global initiatives. Information security, in conjunction with the CSI steering committee, is focused on protecting Cadence source code and our customer IP,” says Rao.