Bandwidth Goes Big With 10-Gigabit
More bandwidth can make backups faster and more reliable, improve communications and data sharing between groups and can allow complex applications to run more smoothly.
Increased bandwidth can improve communications and data sharing between groups. Customers can consider deploying large file transfers or network storage solutions. Voice over IP solutions is yet another network stressor that benefit from network upgrades.
Upgrading also allows complex applications to run more smoothly. Applications generally drive network upgrades because handling multiple tasks and calculations require a lot of network throughput. Take video for example. Even though video servers use multicasting and compression to be more bandwidth-efficient, video that is simultaneously viewed by a large number of employees can strain the corporate network.
Whatever the need, more and more customers are approaching solution providers about installing a 10-Gigabit Ethernet core, or upgrading the existing infrastructure to 10-Gigabit Ethernet. Despite the fact that 10G is just another (albeit, faster) standard for Ethernet, upgrading is not as simple as just using a different network card. There are cabling decisions to be made, new switches and routers to be purchased, and environmental concerns such as space and power to calculate. For this Solutions That Work, the CRN Test Center spoke with Chris Ludwig, the regional director of practice management of Waltham, Mass.-based solution provider NWN, about upgrading the network core to 10G.
Networking vendors have long scrabbled for market share, and 10G products are no different. Of course, the big names, such as Cisco Systems and Foundry Networks, have been delivering 10G products for years, but there are other vendors who've carved out a space for themselves. Force 10 Networks offers both switches and routers in its Force10 E-Series. While Hewlett-Packard's HP ProCurve Routing Switch 9300M Series does not fall in the category of an entry-level device, it's still affordable enough to be used in small deployments. Other vendors include Extreme Networks, Nortel Networks and Woven Systems.
NEXT: THE STEPS STEP 1: DEFINE NEEDS
Just like any other major business undertaking, solution providers and customers have to sit down and have a clear understanding of why the customer thinks 10G Ethernet is necessary. "We need to know, what's the problem we are trying to solve?" said Ludwig. The discussion needs to cover what applications require the bandwidth as well as a timeline to consider how quickly it needs to be deployed.
The discussion is crucial, because what starts out as an application-driven need may possibly transform into an all-encompassing strategy that considers disaster recovery and business continuity. Mirroring benefits from high throughput because the data needs to be written in multiple locations almost instantaneously. A mirror site that contains data an hour older than what is on the main site because of the delay in getting files copied over is not useful to end-users relying on the mirror.
Solution providers need to be prepared to design and deploy a solution that is larger in scope than the originally stated case. Going over the business cases with the customer ensures that needs aren't identified later in the process, since incorporating the new requirements will delay the project. Taking the time to thoroughly understand what the customer is currently doing, and what activities can benefit from the upgrade will save time, labor and costs down the road.
Along with identifying business activities, the solution provider needs to develop a time-based plan that looks at the customer's planned growth. The planned growth considers when the customer intends to add more bandwidth-intensive applications and devices to the network. If it's in the next six months or a year, customers have the option to either "future-proof" their networks, or to opt for a phased installation. By future-proofing, solution providers can purchase and install equipment now as part of the project but not actually hook up the ports until it's needed. Future-proofing gives customers room to grow and expand to prepare for 10G.
Customers may prefer phased installations if they are concerned about downtime. If the company has a lot of work to do for its upgrade, but can't afford to set aside a long downtime window, the solution provider can define phases and identify tasks that will be taken care of in each one. This way, the customer can schedule downtime around its business activities as the upgrade is completed a little bit at a time.
Finally, it is important to thoroughly cover the components that would be taking advantage of 10G Ethernet. Perhaps the company's disaster recovery solution or business continuity sites could be hooked up to 10G. Or perhaps, there is a SAN that that needs to be part of the upgraded network. That segues into step two.
NEXT: KNOW THY NETWORK STEP 2: KNOW THY NETWORK
This is where the solution provider performs a thorough network assessment. Even if the business needs discussion has identified the applications and equipment that will be affected, it is still important to do a full top-down network assessment, said Ludwig. This way, missing and hidden equipment can be identified, and the solution provider gets an idea of how various components are cabled together and where everything goes.
Looking at the entire network, solution providers can help customers figure out if remote closets will be upgraded to 10G eventually, or if the upgrade will just focus on the core, said Ludwig.
If there are duplicate data centers that are far apart, the assessment needs to consider how they would be linked. The site evaluation would also pinpoint SAN solutions, such as from EMC or Network Appliance, using Fibre Channel that would connect to a 10G switch, said Ludwig.
The assessment may include a site survey, interviews with data center technicians and developing business cases. A ROI study or a price assessment is also helpful at this stage in bridging the gap between what the customer needs and what the network can handle. Solution providers study the current network utilization to determine the network's current traffic and usage loads. Various modeling tools can help estimate future loads based on business needs and expected user size. NWN creates simulations based on the customer's network current usage to discover patterns that can be applied in capacity planning.
There are three layers of assessment.
Layer 1 addresses the basic question: cabling and the mode to be used. Copper is preferable to fiber because it takes advantage of existing infrastructure and it is easier to run, but Ludwig points out there are distance limitations. While fiber has more complexity, 10G Ethernet over fiber can go up to 40 kilometers, as opposed to 10GbaseT over copper extending up to 55 meters over a Cat6 cable. The actual distance can vary slightly depending on if multimode is chosen. Some vendors offer proprietary (and even more expensive) optics that can cover distances of 80 to 100 kilometers.
Layer 2 looks at the existing switches and routers that are in place. The assessment considers whether the switches can be upgraded or if they would need to be replaced. Perhaps a simple firmware upgrade would suffice and not require a brand-new unit. "If they have a chassis switch, maybe we can just add a supervisor blade," said Ludwig. The site survey will also identify if the customer already has a blade.
Finally, the solution provider needs to consider the physical space. With new switches coming in and possibly new devices as well, the solution provider should consider the existing power system. The power unit could be upgraded or it could be made redundant to ensure sufficient supply. In addition, a thermal needs assessment and heat analysis would identify necessary changes to the heating and cooling systems.
While this may seem like common sense, rack space is an important thing to consider at this time. As Ludwig pointed out, some customers like to test the new switches in parallel with the existing ones, so additional rack space is an important consideration. Depending on the network's complexity, the assessment could last anywhere from two to three months.
NEXT: ROLLING IT OUT STEP 3: SELECT THE TECHNOLOGY
The network assessed, it's time to draw up a shopping list. NWN generally fills its Layer 1 needs with cabling manufacturers Panduit and Corning. As a Cisco partner, NWN tends to put together a solution with Layer 2 and Layer 3 switches and routing equipment supplied by Cisco. Cisco has a line of solid performance switches that fit in a single chassis. The Cisco Catalyst 6500 Series switch provides up to 64 10G Ethernet ports.
However, this doesn't prevent NWN from deploying non-Cisco products.
"Most companies are using fairly standard stuff," explained Ludwig.
A vendor's commitment to support the upgrade process is a deciding factor in how solution providers select the vendor. Since manufacturers know the products best, the NWN team tells the manufacturer the kind of stuff needed and solicits recommendations, said Ludwig. Some 10G products may hit the market before standards have been ironed out and finalized. In those cases, such as products supporting longer distances, vendors have to offer technical support now and later, as well as the ability to upgrade the firmware to meet standards once they are established. Otherwise, the pre-standard product becomes obsolete. Cisco does this well, according to Ludwig.
Along with the vendor's flexibility, how the vendor handles product failures is also important, whether it's failover capabilities, a responsive product replacement program, failure mitigation technologies, or a product's MTBF (Mean Time Between Failures). Solution providers have to balance cost against uptime promises to make a final decision on selecting a vendor.
STEP 4: PUT IT ON PAPER
The solution provider is now armed with three definite facts. They include the business needs, the current network state and the list of new equipment. At this point, Ludwig and his team draw up plans and design the new network. The diagrams help ensure that nothing falls through the cracks.
Using Microsoft Visio, Ludwig and his team develop network diagrams complete with deployment details. The cabling diagram shows how each of the closets will be connected. The switch diagram shows how the switches will be connected. In the end, an overview network diagram indicates where each piece of equipment would be physically located.
Other documents developed during this phase are labor estimates and project plans.
"The planning is critical in making sure that we don't have time delays or downtime," said Ludwig.
STEP 5: ROLL IT OUT
Implementation should take anywhere from a little less than a month to a full year, depending on the network's complexity. If the design stages, beginning with the business needs discussion and ending with the Visio diagrams, were done thoroughly, the rollout should be relatively free of costly delays or unnecessary downtime. The design stage should have allocated time for a pre-test, when the deployment is tested end-to-end in a controlled laboratory environment. The team would complete the Layer 1 tasks first, to run new cabling and to upgrade existing ones. The cables would be tested to make sure each one works properly. At this point, new power or cooling systems would be installed and new racks constructed. Once the physical environment was ready, the actual switch upgrades and replacements could proceed. After everything has been mounted and connected, there's intensive testing to make sure each component works as expected.
The lab testing is recommended because it irons out many of the issues and kinks that may occur during the actual installation. Delays are costly, especially in terms of downtime. In a dual core scenario, downtime is ideally non-existent, since one side of the network can be shut down while the other side continues to hum along. Once deployment is complete on the first side and tested thoroughly, it can take over essential network functions while the other side is brought down for maintenance.
If the deployment was to support a disaster recovery scenario, testing would include failover, by bringing down one center or the core and seeing if the other centers can keep functioning. Testing is critical. The last thing the customer and solution provider want is to discover a problem with the network during data recovery after a disaster.
Deployment is complete after thorough testing.