How to Setup a Load Balancer with Rackspace Cloud Load Balancers On-Demand

I was sitting around tonight and had an idea. I wanted to know.. How hard is it to setup a load balancer using Rackspace Clouds new offering? I logged in to the Cloud Manager interface. On the left side of the screen I select Hosting to expose the hosting options. I clicked on Load Balancers. Next I clicked on Add Load Balancer. That brings you to the configuration page. This is where you begin by giving a name to the instance. Next choose the Protocol (http, https, smtp, custom, etc..). Then you select the Virtual IP type, there is clear concise documentation that explains the various options. Next you choose the algorithm that the balancer will use. In the next section you select the region for your balancer. I selected DFW because I have several servers (nodes) in San Antonio, and several in various data centers in the DFW area. Finally you add your nodes. The awesome part is that your nodes do not have to be Rackspace servers, be that virtual or physical. You can use physical servers or virtual, and they can be virtually anywhere.. For my test I added a node that was in the Slicehost DFW data center, next I added a node from Linode, also in their DFW data center. Finally I added a node from Serverbeach, a physical server that I lease in their San Antonio data center. Thats it. Your load balancer is setup. Assuming you selected the same options as I did: connection type: http, VIP type: public, algorithm: random, and that you wanted to set this up to load balance all you need to do is go setup an A record in DNS for to point at your VIP (provided to you by Rackspace). Next go to each of your nodes and configure apache with as a vhost. Once that is complete you can visit in your web browser and you are now load balancing! It so simple and only takes a few mins to setup.

5 thoughts on “How to Setup a Load Balancer with Rackspace Cloud Load Balancers On-Demand

  1. I assume if the servers can be anywhere I can load balance my office servers with the rack-space cloud! Must give that a try.

  2. I sat and did the maths regarding RAID the other week for a coenmmt on reddit, based on a 3TB disks being used in RAID.Disk failures happen and worse still the odds are higher that they’ll happen in clusters. If all your disks are part of a batch from the factory, or if there is something quirky about the environment, or something happened during transport, or whatever else, failure of one means your odds of a second failure are significantly higher. RAID gives you an element of protection but consider the rebuild times. Standard speeds are maybe 120-150Mb/s typically. Say 200Mbps tops. At 200Mbps it’ll take over 4 hours to rebuild the array from a single disk failure, assuming the array is idle. Which. of course, it’s highly unlikely to be!If you’re using RAID5 that’s 4 hours of high risk of second failure. RAID6 gives you the extra protection of a second failure, but you’re sacrificing more capacity and performance for the benefit, and increases the amount of data read and written in the rebuild process (and reducing overall performance of the array as you do double checksums)Some back-of-napkin maths: 5 disk RAID5, 3TB disks. Each disk holds 2.4 10^13 bits. 1 in 10^15 bit published error rate for 3TB drives.To rebuild that you’ve got to read and/or writing 12 10^13 bits of data (2.4*5 = 12). At the published error rate that means there is a ((12 10^13 / 10^15)*100 = 12% chance of a bit error occurring during rebuild. On the single parity RAID5 that could be disastrous. RAID6 the advantage of double parity should protect you.Whilst capacities have soared, Mean Time Before Failure has only mildly increased and that bit error rate only marginally improved.I’ve heard of companies that use RAID1 for backup. Each day the admin unplugs one of two drives, and replaces it. With a 3TB drive that’d be a 2.4% chance every day of errors creeping in during the mirroring process alone, let alone normal work load! *shudder*

Leave a Reply

Your email address will not be published. Required fields are marked *