Director vs. Core Switch

Time is money.

At any other time, it’s an overused and ought-to-be-forgotten cliche. But today, it’s perhaps the most poignant and understated one-liner in the business world. Never before has so much ridden on so little, and never before have companies placed such a huge price tag on every minute of every day.

Whether your business is e-commerce, financial services, healthcare, retail, or government, each second your business is running equates to incoming dollars. And, each second your business is halted by data unavailability means lost immediate and future revenue. Even running in a degraded mode can result in prospective customers looking elsewhere

In assessing a business’s application requirements, the application’s current and future cost of downtime must be determined, so the associated network can be designed to support the application’s availability requirements. Generally, if the cost of downtime is deemed high, the need for high-availability solutions is justified.

According to the Association of Contingency Planners, the cost of downtime per hour in the financial services industry is $6.4 million per hour. That means it costs a user $1,777 for every second they can’t access their data. When you look at it that way, it becomes much more evident that availability is the crucial element in the storage networking switch selection process. Relatively new to cost metrics, running at less than full volume can also cost thousands of dollars. Have you ever decided to try a different Web site if the one you are using is sluggish?

When each second of network downtime equates to thousands of lost dollars, companies deploying storage area networks (SANs) today are posed with two choices – Directors and core switches. Directors or core switches are essentially the heart of a company’s data infrastructure, pumping vital information through an extensive and complicated network of cables, storage and servers. If that heart stops, your business flat-lines. Here are a few things to consider when deciding whether to deploy a Director or core switch as the heart of your company’s storage network.

Director or Core Switch?

A Director is a large port count, high bandwidth fabric switch, designed to provide the highest availability and performance. Directors have fully redundant, hot swappable components that minimize downtime. In fact, availability is rated at 99.999 percent or ‘five nines’. 99.999 percent equates to five minutes of downtime per year, the most that a company running applications such as brokerage, online transaction processing or telecom can afford. Also, single-stage Directors utilize a ‘crossbar’ architecture that enables all ports to simultaneously interconnect with each other without performance degradation. This capability of simultaneously interconnecting without performance impact is often referred to as ‘non-blocking’. Directors are a mature technology that complement fabric switches and are used in about 96 percent of the world’s largest data centers.

Core switches, on the other hand, are new to the market. While they offer high port counts, core switches lack many of the attributes that define Director-class products such as highest availability, un-compromised performance, mainframe FICON support and certain environmental advantages. A core switch is basically comprised of a 16-switch mesh. Although easier to develop, a core switch is plagued with the blocking and latency issues inherent to a mesh design.

Availability and Performance

Core switches lack many basic availability characteristics required in a Director-class product. First, enabling new firmware to add desired features or to fix critical maintenance issues requires a disruption resulting in application outages. Second, although processors are redundant, availability is lost waiting for the alternate processor to be activated. And third, a simple component failure on a port card can result in 16 ports (25 percent of the switch) being lost during a service action. With so many ports being lost to a service action, applications and connectivity to critical data will likely be lost.

A fully redundant, single-stage Director benefits from ‘Constant Availability’ with proven 99.999-percent availability in nearly 10,000 installations worldwide. With all components being redundant, failures and service actions (excluding replacing port cards) are completely transparent. Such architecture removes the need to rely on server fail-over software that may have not been tested and verified in some time. Single-stage Directors deliver the industry’s lowest service action impact (4 ports), ensuring applications have the connectivity and bandwidth necessary to meet deadlines and satisfy customers.

And, with a Director’s Hot Code activation technology, critical feature enhancements can be added without affecting critical applications. The firmware of a single-stage Director does not require the Director to be taken offline when deploying new features or applying a maintenance release. This industry exclusive ensures critical server and storage connectivity, resulting in more application uptime with maximum performance. Less sophisticated products such as core switches require devices to be taken offline to activate firmware.

Each 64-port core switch consists of four port cards, with each card containing four switches (total of 16). Within each port card there are two ISL-like paths to each of the other switches. This results in up to 50-percent blocking, and extremely high latency. This can result in sluggish performance and long wait times for applications like online transaction processing or web based retail sales.

FICON

Another area where core switches are sorely lacking is their inability to support mainframe FICON traffic. FICON devices are required to undergo rigorous testing to ensure the absolute utmost in availability, performance and compatibility. A core switch has never qualified as a FICON device. FICON is a widely used, high performance mainframe storage networking protocol developed by IBM for OS/390 class mainframes and is supported today by several major storage vendors.

Directors enable mainframe customers to exploit the addressing, distance and bandwidth advantages FICON delivers over legacy ESCON connectivity. Directors also offer FICON Management Server software, allowing single point network management with SA OS/390. And with recently announced intermix capabilities, customers can simultaneously run mainframe (FICON) and Open Systems (fibre channel) data traffic through a shared Director. FICON Intermix presents the potential for a significant cost savings to customers of all sizes by consolidating data networks, utilizing a single Director where previously two separate networks and much higher costs were necessary.

Environmentals

An efficient single-stage design enjoys significant advantages in power and cooling over a core switch. A single-stage Director requires only 488 Watts, less than 50 percent of the 1,125 watts consumed by a core switch. And in BTUs/hour, a core switch puts out a blistering 3,841 BTUs, over twice as much as a single-stage Director’s 1,673 BTUs. The difference can add up to a significant savings for an enterprise. In a 128-port configuration, the combined power and cooling costs for two 64-port single-stage Directors are $1,894 – less than half of a core switch at $4,350 (assumption of .05 kWh).

Enterprise Legacy

Single-stage Fibre Channel Directors have been shipping for more than two years, with the total number of units shipped and installed more than 8,500 units. Single-stage Directors are installed in 96 of the world’s largest data centers, supporting a multitude of applications in every industry. Conversely, core switches have no proven track record and essentially no data center experience. They were introduced to the market only in the last few months. Core switches have never met the stringent availability and performance requirements of mainframe environments and require twice the electricity of Directors to operate.

While core switches have been thrust into the Director market, they in fact are not Directors. They are a mesh of 16 switches tied together via internal inter switch links (ISLs). They are prone to bottlenecks, failures, downtime and an array of management and configuration problems. So, while core switches are marketed as being comparable to Directors, reality is quite the contrary. And at $6.4 million per hour of downtime, is your company willing to take the chance?

About The Author – Jim Miller is Senior Manager, Product Marketing for Boomfield, Co. based McDATA Corporation.