At Nucleus we can take care of all your server needs. Webhosting,virtual private servers, colocation, consultancy … it’s all available to you. One of our other services is cloud hosting. We often refer to it as a Dynamic Cloud Server (DCS for short). We base our cloud hosting on VMware technology, specifically VMware vSphere. But why would you opt for a DCS? What’s the advantage of a DCS over the old (outdated) physical model?
Many benefits of a DCS are mentioned on our website, but how do they work technically and how do they help you and your company? On the technical side, there are some benefits that we mention but how do they work exactly?
Every DCS has High Availability by default. This allows us to make sure your server remains online in the case of physical hardware failure (the host your DCS is running on). Simply put, High Availability can be described this way: as soon as a physical host fails, your DCS will automatically reboot on another physical host and continue to function as desired. This is one of the greatest benefits of a DCS.
Initially: all physical servers are online..
A physical host fails and High Availability performs the necessary actions:
What if you had a physical server and a hard disk failed? What if the motherboard failed? Hardware failure is something that can always happen and comes with downtime. These can be short but if there’s a big problem with your physical server, it can easily add up to a couple of hours! Can your company continue to function that long without the server your employees or colleagues work with daily?
If you need a new server for your company or environment these are things you need to keep in mind and preferably prevent. A DCS will answer this problem for you: no need to worry about failing hardware ever again. We provide everything to keep your environment online.
How does it work exactly?
VMware vSphere uses an agent called “FDM”, which is short for “Fault Domain Manager”. This agent runs on the physical hosts. Through this agent the hosts can communicate. Besides that this agent also communicates with the tool we use to manage our environment: the vCenter. High Availability makes use of the master and slave principle. There’s always one physical host as the master and the others are slaves. Should the master fail, a new master will automatically be chosen through a selection process.
What is this master responsible for?
- Rebooting a failed DCS
- Passing along information to the vCenter
- Supervising the state of the slaves
The master uses 2 heartbeats methods to know if the host is still available or not.
- Network heartbeat
- Datastore heartbeat
The network heartbeat runs through the master, it will check via the network if a slave is still online or not.
Only when the network heartbeat fails, will the data store heartbeat be triggered. Each shared data store (look at this as a hard disk your DCS is located on) contains a folder with a file called “host-nr.hb”. Data store heartbeat will check the time stamp to see when the file was updated.
If both heartbeats fail, the physical host will be deemed unavailable and the high availability process will start. Your DCS will (if necessary) be rebooted on a physical host that is available.
Of course that whole process happens a lot faster than the time it took you to read this and as a customer you don’t even notice it. You’d only notice in the event your server is rebooted. Your server will become available again within seconds and you can continue working. Being back online that fast would be unattainable if you opted for a physical server, whose motherboard failed.
Besides High Availability there’s an extra way to protect your DCS and that’s the use of Fault Tolerance. This technique will “mirror” your entire DCS to a second DCS. Any change that’s made on your primary server is automatically and instantly executed on the second server. Does the first DCS fail? The second will take over immediately (within seconds) and be available so you can simply keep working.
How does Fault Tolerance work exactly?
As soon as fault tolerance is activated, a ‘shadow copy’ of your DCS is created on another physical host. Neither server will ever be located on the same physical host. As soon as fault tolerance is active, every action you perform will be copied to the ‘shadow copy’. As soon as the physical host your DCS runs on should fail, Fault Tolerance will make sure the server continues on the ‘shadow copy’. This minimizes downtime completely, as your server doesn’t require rebooting, it just keeps running.
Another major advantage of a DCS is the possibility of taking snapshots.
If you need to install a new software version, but you’re not sure whether it will affect your environment, then you can make a snapshot of the server. After this process is completed you can do the upgrade and test everything. A snapshot makes sure that the state of your entire DCS as it is at that moment is stored and that you can roll back to this moment should you have to.
- Does everything work correctly? Then you remove the snapshot and everything keeps going.
- Do problems arise? You can go back to the moment of the snapshot and you’re running on the previous version once more.
When you have a physical server then you need to start restoring backups and you’ll lose a lot of time. Do you know exactly what’s been updated? How certain are you that restoring a backup won’t cause any problems?
Nucleus provides the choice of having snapshot backups. If you chose this we’ll make a backup of your DCS every night and store it for you. To do this we use Veeam Backup and Replication. This software automatically takes a snapshot that contains the entire state of your DCS and stores it on our backup storage. These backups are immediately sent to another datacenter every night. Of course you won’t notice any of this since your server keeps running.
How do snapshots work exactly?
Your server crashes unexpectedly and you can’t restore it? We can reboot your server immediately (instant recovery) from our backups and you’re back online in seconds. This option takes the state of the last backup (snapshot) and uses it to reboot your server.
This is once again a possibility that’s virtually unattainable if you use a physical server.
Placement of your Cloud Server
As we have two data centers available where we can host your DCS you can freely choose whether your server is located in Brussels or in Antwerp. If after a few months you want to switch from Antwerp to Brussels or the other way around, that’s possible. How this works we’ll explain in an upcoming, more elaborate, blog post.