Tips For Building A Redundant Network
by Ed Duncan, Consultant, SBSC, MCTS, MCSE, MCSA
With advancements in technology, many businesses today don’t follow the Monday through Friday 9-5 work schedule. With an Internet presence some businesses are accessible 24hrs a day, 365 days a year. Computer technology makes it possible for even small businesses to offer services and products to their clients at any time. In order for a business to offer this type of accessibility, you need computer systems that are redundant, meaning that if there’s a failure at any point in the network, there are contingency processes in place to keep the network operating and systems available. Building redundant systems needs to be properly planned or the implementation can be costly depending on the level of redundancy you’re looking to build. Let’s look at some of the more critical components of a network to consider when planning for redundancy.
Network Communications
Network communications can be classified into two groups: internal (LAN) and external (WAN). Internal communications will cover your router, switches, and network cards. For redundant internal communication you usually require two of everything. The servers should have 2 network cards, connected to two different switches, which connect to two different routers. If any device should fail, communication can fail over to the other. (Additional configuration is required which we won’t cover in this article).
Redundant external communication most likely is provided and managed by your Internet Service Provider. Ensuring data communications are up, available, and operating at sufficient speed is the ISP’s primary responsibility. If your business has multiple locations many ISP’s provide MPLS (Multiprotocol Label Switching) networks to connect them together over the Internet. The connection from one office to another goes through a mesh network of routers, so if one router should fail, the traffic is rerouted along another path. ISP’s have redundant systems in place to ensure communications are available however unfortunately outages still occur.
Power
There are several scenarios to look at when it comes to power. First let’s look at a single server. To avoid having a server go down because of a failed power supply, you should consider purchasing servers with hot-swappable power supplies. Hot-swappable means the power supply can be replaced while the server is running. The server requires at least 2 or more power supplies, and if one should fail, the others will be able to keep the server up and running. Also, both power supplies should not be connected to the same power source. If the power source should fail that would defeat the purpose of building redundancy.
What if the data center or data closet loses power? That is probably beyond your control to do anything about it. It’s wise to have your servers and network equipment connected to a UPS, but they only provide power for a limited time. Hopefully the building you’re located in has a generator that can provide power to the office for several hours or longer. If not other alternatives to consider are storing your servers in remote facilities, like hurricane proof collocation centers, or use vendors who can store backup virtual instances of your servers, which are activated in the event your primary servers go offline.
Data
Like with power, there are several scenarios as well when it comes to finding redundant solutions for keeping your data available. If we look at a single server again, there are disk solutions available from mirroring, to striping with parity, and a combination of both. Mirroring is the process of keeping a duplicate copy of one drive on another. If the primary drive fails, the mirrored drive can be used until the primary drive is replaced. Striping with parity is the process of writing data across 3 or more disk drives. If one of the drives should fail, the server can still function without data loss. The data stored on the failed drive can be rebuilt once the drive is replaced. If the server has online spare drive available this can usually occur automatically.
Other redundant solutions for data include storing backups offsite, backing up over the Internet, or storing periodic snapshots of your data onto another storage device like an external network drive. This process is similar to mirroring, so in the event you lose a file or have data corruption, you can recover from the snapshot very quickly, without IT intervention in most cases.
Ed Duncan is President & CEO of Duncan Technologies and has over 15 years experience in IT consulting, systems integration, and systems administration, with a focus on IT infrastructure for small to midsize businesses. He can be reached at edduncan@duncantechinc.com.