royal technology management - how cluster works

How The Cluster Works

The cluster uses the technology by AppLogic. AppLogic is the first cluster operating system that is designed for web applications and is optimized for transactional and I/O intensive workloads. It uses advanced virtualization technologies to ensure complete compatibility with existing operating systems, middleware and applications. As a result, AppLogic makes it easy to move existing web applications onto a Cluster without modifications.

The Cluster service consists of several Clusters. Each Cluster consists of up to 32 servers connected by a “backbone” - a 48-port Gigabit Ethernet switch. Each of those servers has one or more hard disks, and two NICs. One of the NICs connects the server to the backbone switch, making it part of the Cluster. The second NIC connects the server to the “front” network, which is connected to the Internet.

Each server that is part of a cluster is called a node, and has a unique name.

AppLogic uses three key terms to describe the things that run on the Cluster. These terms are application, appliance and volume (a.k.a. virtual volume).

Volumes

An AppLogic volume is a virtual storage disk. It has a name, size and data. Unlike physical disks which reside within a given server, an AppLogic volume is a distributed object. The volume consists of a metadata record and at least two mirrors. An AppLogic volume operates exactly like a RAID1 disk array, except that the “disks” themselves are virtual and reside on different servers.

A volume can be resized up or down with a single AppLogic command. It can grow up to the largest amount of storage available on one of the servers on the Cluster, and it can shrink down to the actual amount of data stored on it. For example, if you store 2.56GB of files on a 10GB volume, you can shrink the volume to 2.6GB or so (the inodes, the journal, and other file system data also takes some room). When you need more storage, you can expand the volume to the desired size without

having to copy or migrate files.

Appliances

An AppLogic appliance is a virtual hardware box. Like a physical appliance such as a firewall, a load balancer, or a NAS box, an AppLogic appliance has inputs and outputs and an interface that makes it easy to configure. Each appliance has it’s own operating system and a software stack, integrated and pre-tested for the given function. Each appliance has at least one volume from which it boots. Some appliances have “slots” for additional volumes, so that you can add content, database or other data to them.

Appliances can be as simple as a single Linux box, or as complex as a Clustered database server. An appliance can be started or stopped with a single AppLogic command. Each appliance can be resized - you can choose how much CPU, memory and bandwidth to give it, and you can shrink or expand the appliance later depending on how much load you need it to handle.

Applications

An AppLogic application is a virtual private rack. Like a physical rack, it allows you to assemble and package a Cluster of several appliances and servers, for example, a firewall, a load balancer, several web servers, a database server and a control panel server, so that you can deal with the Cluster as a single box - crate it, ship it across country, roll it into the datacenter floor, hook up the power cable and the network, and you are ready to serve a scalable website or application - all that the customer needs to do is copy their code and data onto the appropriate places in the Cluster.

Each application consists of one or more appliances and one or more application volumes. The application volumes is where your actual site or app (e.g. the scripts/code, the database, the content, etc.) live. You can think of them as removable disks that you can stick into a DVD slot on a given appliance.

One Cluster can run multiple applications. The applications run side by side on the Cluster without stealing resources from each other.

How Cluster Stores Your Data

Cluster stores your data using an approach very similar to the way it would be on a standard dedicated server with a RAID1 array. The primary difference is that with a RAID1 array both disks are actual physical drives, and reside in the same physical server. With CLUSTER, each mirror of your volume resides on a different physical server. The volume as a whole operates just like a RAID1 disk mirror. This level of mirroring protects your data not only against disk failure, but against any failure on the server on which one of your mirrors resides. Each volume is attached to a virtual appliance, which is part of an application running on the Cluster.

This example is for a single Application with 4 appliances being used. 1 for MySQL, 1 for Apache, 1 for DNS, 1 for NAS.

Data written to the volume is sent to mirror 1 and mirror 2 at that same time and is always current. If one of the mirrors (e.g. mirror 1) should fail, the other mirror (mirror 2) will take over. The volume will continue to be accessible, but will go into a ‘degraded mode’ where an alert is flagged for RoyAl Technology Management. RoyAl Technology Management or AppLogic OS will run a ‘volume repair’ operation, which will then restore the disk mirror into the ‘OK’ state where mirror 1 is resynced and brought up to date with the data from the mirror 2 image. This takes the image out of ‘degraded’ mode and back into normal operation. Note that volume repair happens in the background and is completely transparent to your application.

How Cluster provides high availability for your applications

Cluster continuously monitors all physical servers and all appliances running on them. If an appliance crashes (or is rebooted intentionally), AppLogic restarts it immediately on another server. Because of the way AppLogic virtualizes all hardware, the appliance will run there with the same IP addresses and have access to the same volumes as it had before. From the point of view of the software that runs in the appliance, this will look as if a server rebooted and came back.

If a physical server crashes or dies, Cluster will distribute all appliances that used to run on that server across the rest of the Cluster and restart them there. From the point of view of the software that runs inside each of those appliances, the result is the same as if each of those appliances was rebooted. In the case of a physical server failure, AppLogic waits for 3 minutes to ensure that the failed server is not coming back before restarting appliances and declaring any volumes as “degraded”

The whole process is completely automatic, and requires no human intervention. In most cases, by the time the RoyAl Technology Management receives the alert about a failed server or appliance, all affected appliances are already up and running elsewhere on the Cluster. What’s left to the operator is to take care of the failed hardware (if any) by replacing it with another server, or repairing it. Once the failed node is back online after repairs, it is now part of the standby resources ready to take over if another server fails.

Overall: It is a system that has fail-safes to keep everything up and working for you.

Below you will see a graphical representation of this (click on image for larger view).

By using RoyAl Technology Management’s services you agree to the Acceptable Use Policy and Terms of Service