XILO Communications Ltd. : Blog

Support 0333 773 3000 • Sales 0800 905 5000

BLOG

Our blog includes company news, recent announcements and upcoming service changes.

High Availability Hosting and Virtual Environments at XILO

You’ve probably seen the term cloud hosting banded around. Many companies now use it to keep your storage ‘in the cloud’. However, how is this different from before when they’d store your information, such as emails and files you have received to information and data you have uploaded? The answer is, it’s not very different at all.


Cloud Hosting?! What if it rains?

The information is still stored on an internet connected server (or servers), somewhere in the world. It’s all ‘in the cloud’, as it’s all on the internet. On network diagrams, the internet has been depicted as a cloud for many years.

blog image


Uptime for us, is one of the top priorities in providing services to our clients. Over the years, downtime has been minimal with our yearly uptime equating to over 99.99% across all of our services (this equates to 52 minutes of downtime a year).

Many of our servers use a control panel such as cPanel or Plesk which isn’t scalable at all. You can farm out DNS records and authoritative servers to alternate locations, which we do, but this is no use if the actual server hosting the content, or mailboxes for example, is having issues. We have even setup secondary mail servers to capture mail in case the primary mail exchanger (MX) is down. This will store and forward the mail when the primary comes back online without generating temporary error messages to the sender.

In the event that a server needs to be rebuilt, it can take time and this can lead to sites not working until the server is restored to service. In our 6 year history, we only had to do this once or twice which is testament to how reliable our server hardware and platform are, as well as how stringent our security is.


Which Solution?

We’ve reviewed nearly every solution under the sun, to match a set criteria for the service we want to offer. We currently use OpenVZ, which works very well for us and we will continue to offer it for standard virtual servers.

In addition to this, we have also watched the industry to see what others are doing, how they perform and what problems have been encountered.

Xen appears to be a popular choice amongst many companies and when used with SAN storage, can provide near instant recovery to another node within the same grid. Scaling your application would simply be a case of deploying another node to the grid which could be time consuming if there are multiple nodes that you need to configure each time you expand.

Virtuozzo is the paid-for version of OpenVZ with several improvements in recent releases not included within the open source version of VZ. At the time of looking at this solution, there was no storage apart from the machine that housed the machines so fail over would take time and would depend on the host node being available.

OpenVZ for the same reason as Virtuozzo was ruled out from our ‘high availability’ solution.

VMware is also another popular choice and has proved to be a very reliable and scalable package but support during our time using the software, as well as trying to get pricing and other implementation information was not readily available which was a major concern.


What does this provide?

Our cloud allows us to provide hosting solutions with high availability built right in to the core of the service. It also allows us to scale the appliances across many nodes which could total hundreds of CPU cores, gigabytes of memory and storage space. Much more than can be offered with services from other vendors using Xen or VZ single-nodes.

Many popular software services can be built on the platform in such a way that should a surge in demand for your application, site or service occur; resources can be added very quickly allowing you to deal with such a surge.

This also delivers true SaaS (software as a service) including helpdesk and CRM applications such as Cerberus or SugarCRM. It’ll be hosted on our network, in the cloud and data stored on our fully redundant IP SAN. If one of the nodes fails, any affected part of your application is started elsewhere without any manual intervention.

We will also allow us to offer some of the first virtual appliances with IPv6. We already have a large IPv6 range allocated by RIPE and are looking to start testing this at a later stage which is something we cannot offer with our existing VZ based solutions.

Over time, we will be testing and looking into multi-site database replication for ultimate peace of mind and additional ‘cloud’ locations such as Manchester, Amsterdam and America.