Network design for a data center is more than just making a bigger network than what you have at home. It involves creating a plan based off of what matters the most to you or your business and selecting gear that brings that plan to life.
When it was time to design SingleHop’s network, we had a lot of considerations in mind. We needed our network to have high availability with 100% uptime, we needed core redundancies in place, and we needed reputable providers.
Three main factors went into designing our network:
We built our network around our SLA to make sure this super network supported our 100% uptime promise with concurrently maintainable systems. We have two core platforms, which connect our network to the Internet. Each Tier provider connects to our data center with two connections from different physical entry points in to the data center, one for each core. The connecting paths are diverse and help prevent downtime due to any potential physical damage. Any failure of the core could affect most of the data center for a long time, but cores rarely ever die completely. And that’s why we have two.
Aggregation is another component. We have a redundant aggregate layer, which splits the Tier lines. We invested heavily into our aggregation because failed aggregation could adversely affect hundreds of customers for a long time.
A cabinet (CAB) switch is another element in our network design. We have one switch per CAB; all the servers are connected to it and it is connected to the redundant aggregate layer. If the CAB switch ever fails (a rare occurrence), it is easily and quickly replaced in about 20 minutes.
As mentioned above, the Tier providers connect the data center cores to the Internet. Generally, you have three Tier level choices to develop your upstream strategy. A Tier 1 network is similar to the backbone of the Internet with the best quality connection. Tier 2 is generally a blend of a Tier 1 quality connection and Tier 3 reselling services. Tier 3 is typically resold IP transit (usually from Tier 2 networks) and, quite honestly, we wouldn’t be caught dead using it. We use Tier 1 providers because they’re the best for supporting our customers’ needs.
When buying network gear, we wanted to choose a provider that could support our network, upstream, and SLA. We also wanted a brand that could ensure top notch security within the network. We went with Cisco because they are the data center industry standard and are very easy to automate and work with. Another great provider is Juniper, who is right behind Cisco in market share. Recently, we’ve started adding more Juniper devices to our network because they’re easier to work with and have some great tools focused around multi-tenancy.
After we rounded out our choices, we had to put it all together. We also had to figure out how to migrate the network we already had over to this super network. The cost was obviously going to be high, so we figured out where we needed to invest the most money, choosing to invest more in our core routers and aggregate layers and a bit less into our CABs.
It all worked out and now we have 100% uptime everywhere. We have this high availability network because we bought two cores to support our clienteles’ needs. SingleHop’s network is successful because we invested so heavily in our equipment and connections in the early stages of its design. Now we can take down half the network for maintenance and no one will have any downtime, which we did just the other day! We wanted to build a top notch network to keep our customers working 100% of the time, and I really think we achieved the best network possible.