Clustering for High Availability and Failover
Clustering server and computer systems is a very important design consideration when planning out your network. As a systems integrator with decades of combined experience in IT Service, Aertight has designed Microsoft Windows Failover Clusters for everyone from fortune 500 companies to small local datacenters.
Clustering works by taking several powerful computers and linking them together with software, so that they each system is "aware" of the states, resources, and "availability" of the other systems in the cluster. When clustering several systems together, it's common to host either full virtual machines or individual applications on a SAN, NAS, or other shared storage appliance.
Clustering gets complicated, fast.....
So the old phrase is, what a cluster-%#@! , and it's been exclaimed by even the most professional IT services personnel after 4 hours of trying to track down an issue in a cluster. In essence, clustering gets complicated because it needs to offer failover and high-availability to the virtual machines and applications it hosts, while maintaining a level of redundancy for these services. In order to achieve this, you really need a system that is "N+1", which means you have N parts in the system, and you always have +1 parts available to the cluster if one of the "N's" fails.
Practically speaking, this means you're doubling up on "everything".... Double the hard drives, double the network cabling, double the switches, double the shared storage, and everything else you need to make clustering function. It really doesn't make any real difference if you use Microsoft, Linux, or Bob's Discount Operating System, you're still going to have a ton more equipment and stuff to manage. On top of that, networks don't like loops, and even with the latest queuing and switching protocols, you can still make a mistake, and make your life a total nightmare in trying to entangle a spaghetti mess.
Aertight built a better clustering system
So, we're engineers, and we talked with other engineers, and we decided we would build a system that could really help solve many of the issues with traditional clustering systems setups. We listened to our clients, and they told us they wanted a system with network mirrored (not shared storage) between cluster hosts, and they wanted to ensure the entire cluster was N+1 redundant at the "box level". In short, clients cared less if they had a single hard drive on a single host in the cluster that was down, rather than having a failure of the entire cluster because of something remote to the hosts (usually the networking, switches, cables, ect).
Our cluster solution is called the Aertight Failover Cluster System
The Aertight Failover Cluster System combines two Aertight Server Platforms (our all-in-one rugged, fanless, water-resistant, dust-resistant, touchscreen embedded, Microsoft Windows Server System), and connects them with two passive cables. Using network disk mirroring technology, we avoid the main sources of issues in the cluster, including all the N+1 network hardware. With our failover clustering system, you only need 2 passive cables between the servers. The systems themselves then mirror the shard data across high speed network connections. Microsoft Windows Failover Clustering can then be used to build a cluster that's redundant and reliable, without all the extra complication and cost of the rest of the network gear. We build the Aertight Server Platforms like a tank, so we're confident they'll be the most reliable best clustering system you could implement.
Would you like to learn more about our clustering system?
