HIGH AVAILABILITY WEB HOSTING

If you're new to web clusters or high availability hosting, please take a moment to read up on what we do, and why we're here! This article will help you understand the difference that web clustering makes for your business.

What is a web hosting cluster?

Clustered Hosting, according to Wikipedia, is "having multiple servers hosting the same content for better resource utilization. Clustered Servers are a perfect solution for high-availability dedicated hosting, or creating a scalable web hosting solution. A cluster may separate web serving from database hosting capability".

Our architecture is a true web-cluster, and it was designed specifically with the key concepts in mind as described in the definition above. To take the definition a step further, we like to consider a "cluster" a group of servers that have been configured in a way that they all work together and look and behave like a single server. At a very basic level, our web cluster acts just like any "single server" web hosting system - and you can treat it as such - with these added benefits:

• High Availability: Since a "shared" hosting account on our high availability cluster is actually served by a network of 8 or more servers, we can provide improved reliability and "uptime" over a single-server solution. Every system is redundant; so if one of our web servers goes down, then load balancers automatically recognize the failure and take that "node" out of the web server pool.
  
  
• Scalable Hosting Platform: One of the major limitations of other shared and even dedicated server hosting platforms is that once you hit your server performance peak, then only option is to upgrade your hardware or add a new stand-alone server. In a typical shared hosting environment, this means potential downtime and risk to do hardware upgrades. In a web cluster, we can add capacity instantly by simply adding a new server to the cluster. We can even take a node down for hardware upgrades, and your site will continue to run on the rest of the servers.
  
  
• Better Resource Utilization: Another common issue in the shared hosting environment is the case of "lop-sided" utilization of servers. Often as a hosting company grows, they will add more single- server installations, leave the old clients on the old servers, and begin adding new clients to the new server. After adding a new server, the new and usually most powerful server in the network sits doing almost nothing, while the older servers in the network are doing the bulk of the work for older clients. With a cluster (and with our load balancer), we can immediately take advantage of newer and more powerful hardware, and ALL clients immediately get the benefits of the new server.
  
  
• Server to Service Tuning: As mentioned by Wikipedia, a cluster may separate web serving from database hosting capability. What they don't say is this means we can tailor our hardware and software installations to match the system need. Database servers require much more memory than a typical web server, but when using MySQL can only take advantage of up to 8 processor cores. So a web server might require slower processors with more cores and lower memory (for example) than a database server, which needs the fastest 8 cores you can get with tons of fast RAM since you can't add capacity by adding more cores..

What is load balancing, and why is it good for my site?

Load balancing is an essential component of a well designed web hosting cluster. Load balancers simply spread the load of all of the web requests across all of the available web server nodes. Essentially the job of the two load balancers are to receive ALL requests for all web sites hosted on the cluster, and forward these requests to the least loaded server node in the cluster. The load balancers also monitor one another AND all of the web servers in the cluster. If the primary load balancer fails, the secondary picks up within seconds of the failure. If a web server node fails, the load balancers will remove that node from the "load balancing rotation", so customer web traffic is no longer sent to the degraded or potentially dead server node.

What does high availability hosting mean?

High availability hosting is a method to eliminate as many single points of failure in a hosting network as possible, in order to provide the best uptime possible. While there are many ways to provide high availability we have chosen to do so using a custom built web hosting cluster. In a nutshell, if you're on a single-server hosting solution and the server goes down - scheduled or unscheduled - your website goes offline. This could be anything from a RAM upgrade to a catastrophic hard drive failure.

In a high availability cluster solution, both the hardware and the services are redundant. Nearly any single component could fail and the network is resiliant to the failure - the websites continue to run. We offer the following redundancies in our HA network:

• HSRP Network Connectivity: Dual network feeds from our data center provider (also know as HSRP connectivity) provides network level redundancy to our entire network. In the event of a network card, switch or router failure at our upstream provider, traffic is automatically routed to a hot standby router and connectivity is not lost.
  
  
• Hot/Standy Load Balancers: We also provide an internal HSRP style connection from our load balancers to our web servers. In the even one load balancer fails, the hot standby again immediately picks up right where the primary unit left off.
  
  
• Load Balanced Web Servers: We have multiple load balanced web servers in the shared hosting cluster, instead of a single machine-per-client as is typical for most web hosts. Since your web site can be served from any web server in the cluster, the failure of a single server does not cause any down time or create signifant loading issues on the network. We will maintain sufficient reserves in processing power per-server such that any single web node could be lost without causing any significant performance issues across the network.
  
  
• EMC RAID 5 Storage: RAID stands for "Redundant Array of Independent (or inexpensive) Disks". What it means is that any single hard drive could fail in our (expensive) EMC RAID array without any data loss. RAID 5 is optimized for general purpose speed, throughput and fault tolerance. And Big Iron RAID boxes are far from "inexpensive".
  
  

Does my application need to change for High Availability?

In most cases NO! We utilize SAN storage with a clustered file system (Oracle's OCFS2 Cluster File System), and put php sessions and private data files directly on the clustered file system. This means that each web server can treat sessions as local to the server, while allowing the session information to be shared across the cluster. So it isn't necessary to make your applications cluster-aware - it just works! However, if you'd like to take advantage of load scaling to mutltiple read-only database servers, let us know and we'll create a customized solution just for you.

Do I need to worry about privacy on the cluster?

We worry about that for you. Like any shared hosting environment, privacy of both scripts and data must be treated with care. We have taken the utmost care to ensure all of your data is "sandboxed" into a custom private environment. We run PHP always in safe mode, though due to the method of installation, most applications that claim to need safe mode off can run on our cluster. We run a customized SUexec for CGI based applications such as Miva Merchant and Perl to ensure that all applications are sandboxed into the owners home directory. And we've written custom extensions to several opens source applications such as vsftpd and openssh to ensure that FTP and SSH based access is also compartmentalized.

Do you provide email hosting in the cluster?

No - our cluster servers are web application servers, not email servers. Truthfully, you shouldn't be running your company email on the same server as your online store. Google Apps has an excellent FREE email application for up to 10 email accounts, and there is a minimal fee if you need more than this. WE DO NOT PROVIDE EMAIL SERVICES excluding store based outbound emails for invoices.

Can we send email marketing from our cluster nodes?

No - our cluster servers are web application servers, not email servers. Please see our anti-spam policy, and sign up for one of the great email marketing services such as ConstantContact, MailChimp, etc.

Redundant hot/standby load balancing routers for High Availability.

High performance next-generation clustered file system - no NFS here!

Fibre Channel RAID 5 storage for maximum reliability and speed.

Tier 1 Data Center facilities with N+1 redundancy from cooling to switches

Custom hosting control panel designed for clustering.

Custom cluster design for colocation of highly available networks.

Reseller and white-label programs available.

Dedicated Nodes include shared cluster "fail over"