Fabric-based computing

This technology will enable data center administrators to use centralized management capabilities within a unified domain that serves as a central nervous system for unified computing. By Nivedan Prakash

With the advent of Fiber Channel over Ethernet (FCoE), fabric-based computing is starting to enter the mainstream of enterprise computing. It is a method of unifying the network—both FC and Ethernet—in data centers. FCoE is the communication protocol which enables this unification. Its benefits range from elimination of dual networks—FCP and Ethernet (read cost reduction)—to simplification of the host-to-fabric connectivity.

According to industry experts, fabric-based computing is a concept, which helps companies introduce a ‘wire once’ deployment model. With this computing model, although the configuration changes, companies will not have to install new components or re-cable existing ones.

A unified computing architecture gives data centers room to scale while anticipating future developments in technology. Unified computing elements are pre-engineered to accommodate upcoming technologies such as 40 Gigabit Ethernet when it becomes available. The architecture's simplicity allows data centers to scale in size, performance and bandwidth without the historic added complexities of older platforms. This approach helps increase ROI while protecting that investment over time.

“Fabric-based computing is the next-generation architecture for enterprise servers. It combines powerful server capabilities with advanced networking features into a single virtual server configuration. In fabric-based computing, resources are no longer tied to a specific physical machine. They can be reconfigured on the fly without adding software layers. Everything is done at the hardware level (CPU memory, network I/O, etc.). Reconfiguration doesn’t happen at the individual machine level but at the ‘fabric’ level,” said Sundararaj Subbaravalu, Partner - Convergence, Healthcare and Technologies and Founding Team Member, Anantara Solutions.

The availability of a large number of compute, storage and connectivity components around this unified framework makes it an interesting trend to watch. With data centers being run over capacity, there arises a need for unified architecture, which can consolidate the network, storage, and compute components over a unified fabric.

The scene in India

According to an IDC report, the Indian data center services market is expected to touch around Rs 10,000 crore by 2011 at a CAGR of 22.7% over the two-year period, 2009-11. The overall India data center services market was estimated at Rs 6,300 crore in 2009. The potential is high and adoption levels in compute-intensive industry verticals such as pharma, insurance, consumer banking and hedge funds are bound to increase.

Naveen Mishra, Principal Research Analyst, Gartner, was of the view that fabric-based computing is a niche right now. The Indian market has seen a few early adopters, especially some of the large data centers looking at a fabric-based environment for different workloads but not necessarily for mission-critical workloads. The early users of this technology are large banks, financial services companies and some manufacturers. From the end users’ perspective, it has been long awaited as it provides a great level of flexibility.

Industry watchers believe that many large enterprises in India are either evaluating or have implemented pilot projects around fabric-based computing. The biggest draw is seen around virtualized environments. The biggest growth in India would be visible once the stability and acceptance of fabric-based compute frameworks is established around enterprise applications. The ramp-up is expected over the next 12-18 months.

Sumit Mukhija, National Sales Manager - Data Center, Cisco India and SAARC, pointed out that the vendor had received an overwhelming response from the Indian market. Fabric-based computing was, he felt, the best mechanism for data center consolidation and server optimization. The growth drivers for this model in India are virtualization along with cost advantages.

According to Mukhija, the benefits of virtualization are accepted and recognized by organizations across verticals. In recent years, the majority of IT departments around the world have moved beyond data center and infrastructure consolidation to virtualization. However, despite the widespread adoption of virtualization, data center operating costs are at an all-time high. The next-generation data center needs to be viewed across all the discrete technology domains as a single, integrated design. This can be achieved by adopting a fabric-based computing model.

The advantages

Fabric-based computing helps reduce cost by eliminating the need for a parallel set of components to support multiple data transmission protocols. Along with the reduction in these infrastructure components on the server side, a similar reduction on the network edge is possible, further reducing costs.

Aggregating the server's I/O resources saves significant capital expense. Consolidating resources over the unified fabric eliminates the cost of underutilized Fiber Channel HBAs and NICs as well as associated cabling complexity. Instead of being designed to accommodate bandwidth peaks using a dedicated switch port for each host, a data center can share remote Fiber Channel and Gigabit Ethernet ports, enabling network designs based on average load across multiple servers. This can save up to 50% of the cost of the I/O associated with a server.

“Unification of data transmission protocols coupled with the unification of data management processes has long been adopted by the industry. A parallel set of components induces cost increments both during acquisition and during the life cycle of data management that follows. Cost reductions achieved via the elimination of redundant components are amongst the first objectives of every CIO today. Although there are data centers that still use parallel sets of components, their numbers are dwindling. Recent advances in technology are helping data center operators cut cost by improving overall efficiency and RoI mostly by achieving the unification of data transmission protocols and data management processes for backup and disaster recovery,” asserted Satyaki Maitra, Director - SI Business, NetApp India.

Maitra added that fabric-based computing would usher in the next level of unification, this time at the network layer, by eliminating two different layers of network namely Fiber Channel and Ethernet hardware. The ripple effect in cost reduction is witnessed both in acquisition and in operations. Large adoptions will tilt the cost model further towards a shorter RoI.

Moreover, the fabric-based computing architecture can deliver up to four times the compute and four times the bandwidth capacity in the same footprint, up to 92% fewer points of management than legacy networks, up to 30% greater application throughput with the Virtualized Interface Card, up to 76% database consolidation with memory extension and a reduction in power consumption of up to 10%.

These results translate to reduced infrastructure complexity and sprawl, significantly reduced costs and improved business agility. The solution delivers end-to-end optimization for virtualized environments. It is built to meet today's demands while being ready to accommodate future technologies including more powerful processors and faster Ethernet standards as they become available.

Enabling ‘wire once’ deployment

This technology enables a ‘wire once’ deployment model where changing configurations no longer means installing new components or re-cabling existing ones. The concept of a unified fabric is to virtualize data center resources and connect them through a high bandwidth network that is scalable, offers high performance and enables the convergence of multiple protocols onto a single physical network. These resources are compute, storage and applications, which are connected via a network fabric.

According to Mukhija, Cisco’s Server Fabric Switch creates a unified, ‘wire-once’ fabric that aggregates I/O and server resources. With the unified fabric, instead of servers having many cables coming out of them, the server switch connects every server with a single high-bandwidth, low-latency network cable (two cables for redundancy). This setup aggregates Ethernet, Fiber Channel, and clustering interconnects into a 10-Gbps InfiniBand cable. The server switch then connects servers to a pool of shared Fiber Channel and Ethernet ports over line-rate gateways and creates virtual I/O subsystems on each host, including virtual HBAs and virtual IP interfaces. Servers can then share a centralized pool of Ethernet and Fiber Channel ports that can be upgraded and serviced without affecting running applications.

“With the meteoric rise of adoption of fabric-based computing, terms like ‘wire once’ are going to become common terminology very soon. Clearly the use of such terms will help simplify the understanding of benefits around this framework. Caution must be exercised to avoid over simplification. The right expectations and right solutions would ensure the success of early adopters and hence the final success of this framework,” said Maitra.

Different approaches

In the recent past, Indian service providers have made significant investments in the area of virtualization and now the market is observing initial vendor investment for supporting fabric-based architectures. This is because service providers are always looking to adopt technologies designed to reduce cost and fabric-based computing is being watched as well as evaluated with great interest. Although large deployments are still being planned, many SPs have already adopted the framework for smaller application landscapes with their data centers.

Most IT service providers are increasingly investing in this space as the concept allows organizations to increase effectiveness by maintaining IT disciplines and accountability while increasing teamwork and collaboration. Server, network and storage administrators can preserve accountability for their domain policies while interoperating within a single integrated management environment. Computing infrastructure can then be provisioned without the clumsy time-consuming coordination required by legacy infrastructure. As data center roles and policies evolve, individual responsibilities and system privileges can be easily modified and new roles quickly created.

“As per the latest indications from the market, 40% of Indian enterprises are in a state of readiness to move to a newer platform, either completely, or partially. This is in stark contrast to the mere 20%, which were interested a year earlier. The primary reasons for increased fabric-based computing adoption are efficient allocation of resources, resource pooling and distribution, liberation of resources to focus on core competencies and, lastly, it helps SMBs embrace this architecture on a pay-per-use mode,” opined Subbaravalu.

Every technology provider either already has a product or intends to release one in the near term. This technology is not a point solution but a ecosystem of correctly tuned, tested and productized versions of multiple products. Moreover, the selection of such solutions that scale to offer stringent SLAs required for enterprise data centers will ensure success and help achieve RoI in a shorter timeframe.

Coming soon

According to industry experts, fabric-based computing is expected to become a reality for most enterprises in the near future. As a result of the virtualization revolution of the past few years, new standards, technologies and integration conventions are emerging. These innovations enable the design of a pre-integrated data center solution from industry-standard components, rather than requiring IT staff or consultants to integrate their own network, computing and virtualization platforms to create a data center solution.

In order to adopt this market transition, organizations need to embrace an architecture which uses unified fabric and provides transport for LAN, storage and high-performance computing traffic over a single, cohesive infrastructure. This approach can consolidate or entirely eliminate multiple server adapters, chassis switches, cables and other supporting infrastructure. This simplification can also reduce by half the supporting infrastructure requiring power, cooling, management and security as compared to traditional computing environments.

[As with many new supposedly disruptive technologies, fabric-based computing faces resistance largely because adopting it necessitates replacing a lot of the network/switching fabric in an enterprise which could prove expensive. While the benefits are tangible, the initial hump of CAPEX has to be crossed to get to them. - Editor]