(WEB HOST INDUSTRY REVIEW) — Operating a shared hosting business has always been a kind of balancing act, in which the host strives to get as much use out of its servers as possible, while still providing a reliable level of performance and service to its customers.
Over the years, tools have emerged to help hosts with the process of optimizing their operations. And while those tools have taken many forms, CloudLinux (www.cloudlinux.com) may be the first operating system designed with the shared hosting environment in mind.
While operating systems has certainly involved over the last decade to account much more for the functions of hosting servers, CloudLinux, launched earlier this year, is designed specifically to operate in a way that benefits shared web hosts, by splitting server resources into virtualized environments.
Transitioning to CloudLinux is simple, and the performance benefits are immediately apparent, according to CloudLinux CEO and founder Igor Seletskiy.
In a long and detailed email Q&A with the WHIR, Seletskiy outlines some of the objectives and benefits of the CloudLinux operating system, and explains how its Lightweight Virtual Environment technology compares to some of the other hosting architectures out there in the market.
WHIR: Can you explain in some detail the role Cloud Linux is intended to fill in a hosting environment? Obviously it’s an alternative OS, but it also has functions around virtualization and some related to security. Can you break down what a host might be replacing by deploying Cloud Linux?
Igor Seletskiy: First and foremost, CloudLinux is an OS, so, the only thing a customer would be replacing is their OS. It is an easy and straightforward switch from CentOS or RHEL to CloudLinux. All other components stay the same: control panel, virtualization, and backup – we made sure we work with all major vendors and that our OS is compatible with their software.
Second, we understand the Web hosting industry has needs and requirements that are very different from typical needs of enterprise or desktop users. Web hosts and data centers require better control of and efficiency from their infrastructure if they are to be competitive and profitable. And it’s not just to provide better control because web hosts require specific capabilities that enterprise and desktop customers do not need. At its core, CloudLinux was developed to “crash proof” shared hosting infrastructure providing increased stability and infrastructure reliability since most other solutions tended to be too “generic.”
Let’s take “security” for example: in enterprise settings SELinux is widely used, yet in the hosting industry, SELinux is disabled at almost all companies. It just doesn’t fit the need. Stability is another thing that takes different forms for hosting companies versus enterprises. Enterprises, for example, know exactly what they run on their server, and can fully control what is executed on the server. All they expect from the OS is to provide a stable base for their applications. Yet, in hosting settings, the server owner has little control over the applications being run, and as often happens, one application can adversely affect other applications. An OS that can limit this negative effect would be very beneficial.
One can see therefore, that the industry needs a stable OS, optimized for security, performance and stability of these very specialized needs. Within this specialized space, we created the CloudLinux OS so that the only thing customer would be replacing is their OS. It is fairly straightforward switch from CentOS or RHEL to CloudLinux. All other components would stay the same: control panel, virtualization and backup, since we try to make sure we work with all major vendors, and that our OS is compatible with their software.
In the end, we focused on solving a problem that is unique to shared hosters and data centers, yet executed within a straightforward OS platform that helps improve stability, security and performance for anyone doing hosting.
WHIR: Let’s talk specifically about the virtualization component. Can you discuss Cloud Linux virtualization in comparison to other existing forms of virtualization that might be in use at a hosting provider?
IS: On the market today there are two major virtualization approaches – hardware (hypervizor: xen, kvm, vmware, baremetal), and OS (container based virtualization: virtuozzo, solaris zones, freebsd jail).
Hardware based virtualization is the highest level virtualization with most overhead. Each virtualized environment runs its own, fully featured OS. It means that everything, including hard drives, is virtualized.
Container based virtualization shares the OS itself, but creates containers and applications inside one container. One container cannot see, talk to or affect applications in another container. With container based virtualization, the kernel, process, memory & IO management is all shared, so there is much less overhead than with hardware-level virtualization. It also means there is less flexibility since all containers share the same kernel, modules, etc.
CloudLinux provides “application” or “tenant” level virtualization. It is somewhat similar to OS based virtualization, though we virtualize the application itself, not the OS. This means that while there is one application (like web server), each tenant can only see its own files/processes and cannot affect other tenants (CloudLinux limits the amount of CPU each tenant can get).
Here is what it means at a practical level. If you have a server with 250 sites, and you want each site to be virtualized, with hardware and OS virtualization you would need:
1. Each site would run on its own virtual server
2. Each server would have root access, require regular updates, and would need a sys admin to operate
3. A separate installation of the web server, control panel, and backup software.
4. Separate config files
The result is you would have 250 servers, 250 control panel licenses, 250 web servers, and it would probably not fit on high end 8-core servers.
Yet, with application level virtualization, you can still:
1. Have one server to maintain
2. Have one instance of the control panel, web server, and backup software
3. Have a single config file
4. Run it as you run today
And a regular dual CPU server should be able to handle it with ease.
Application level virtualization is very close to OS level virtualization, yet it works on the application level itself meaning that anything outside the application is not virtualized. Yet, inside the application, CPU and filesystem resources are virtualized. This greatly improves stability and security, as it prevents one tenant within the application from affecting another tenant.
WHIR: Is there a certain kind of hosting product, or size of customer, Cloud Linux is designed around? The built-in functions around virtualization would seem to become less relevant in a situation where the host is selling dedicated servers, for instance.
IS: Our sweet spot is shared hosting providers. Someone hosting anywhere from 20 to 10,000 sites per server would find what we offer irresistible. It is like driving the same car, yet getting double the mileage and one tenth the problems.
CloudLinux has benefits for someone who is running a dedicated server for one or two sites but it is not as straightforward. Security is one benefit. We virtualize the site, which means the site cannot affect other sites. It also means that it is much harder for the site to affect the server itself. If the site is hacked, all the hacker can do is alter the content of the site, they cannot touch the configuration files of the server.
Also, since the majority of sites are represented by multiple web applications, we allow easy segregation of web applications, so that one web app cannot affect the other one. Previously, it would be necessary to get another server (or VPS) to do that, which then creates lots of extra hassles, costs and inefficiency. With LVE the whole process is much easier and doesn’t create the overhead associated with running multiple servers.
WHIR: Let’s say a shared or VPS hosting provider is already operating using pretty standard components. How complex would the process of migrating to Cloud Linux be?
IS: It is usually around five minutes per server, as long as hosting provider runs CentOS 5.x or RedHat EL 5.x. It can be safely done on production servers, and all our customers are upgrading their existing servers. It is never about migrating customers from one server to another.
The process has two stages: run the script that converts CentOS server to CloudLinux server (one to two minutes, plus a reboot). Then run a script that installs extra apache modules depending on control panel that you are running (another one to two minutes). If you are running LiteSpeed webserver you can omit step 2. It will detect CloudLinux and start virtualizing tenants automatically.
People running R1Soft backup would need to re-run r1soft-cki.
When we developed CloudLinux, it was critical to ensure that migration to our solution would be easy and fast. The experience of our customers confirms that we succeeded in this goal.
WHIR: Supposing a host were to make that transition – again, lets imagine it’s from pretty commonly used OS and virtualization components – what kinds of aspects of their environment or their business’s operations should they expect to see improve, and how?
IS: If you are a shared hosting provider – you will get stability improvements right away. Your system will run more smoothly and you will rarely have customers calling about their sites being slow or server not responding.
This is the key for our customers and the majority of them start with their “problem” servers. They install it on servers where site responsiveness is slow from time to time or it simply goes down. And the problem often is quite evasive which makes it hard to pinpoint. These are the types of conditions where CloudLinux makes a noticeable difference and we acquired quite a few customers that way. They install CloudLinux on one server, run it for few weeks, and start installing CloudLinux on all other servers they have.
Next thing hosting companies realize is that they get much better visibility into what is going on with the server. They see actual usage patterns on a per customer basis. They see which users use majority of resources, and should be upgraded to VPS. Overall, this provides superior control and insight into customers’ usage patterns, which could help identify new revenue opportunities.
Finally, customers can expect to benefit from increased density. As hosts get better understating on how to utilize their server, and the server gets more stable, they are able to add more and more clients.
Today, the majority of servers at shared hosting companies are utilized at 10 percent to 30 percent of their capacity. This is done to make sure that servers have some room to deal with usage spikes. Hosts can easily double the utilization without sacrificing stability (as CloudLinux prevents spikes from one customer affecting other customers).
With new SecureLVE that provides filesystem level virtualization, we will also radically increase security of shared hosting setup. By segregating each user into its own virtualized filesystem, we are making sure that even if one customer is hacked, hackers cannot explore and modify content of the sites of any other tenants on that server.
There are a myriad of improvements a hosting provider can benefit from as soon as they make the transition to CloudLinux.
WHIR: You’ve only been in the market for a short while. What kind of response have you seen from the hosting community up to this point?
IS: The feedback from our customers is great. They love the improvements they are getting. Given that they didn’t have to change the way they operate, that means a lot to them as well. We are getting great support from various software vendors that are playing in this market too. And the proposition of what we offer is very clear to control panel makers.
Yet, I am afraid we have created some confusion with our name since people may think we are some kind of cloud, server virtualization or clustering software provider. I like to explain that we view ourselves as a building block for vertical clouds, where application/tenant level virtualization and high density are very important. CloudLinux is crucial for web hosts and data centers because it provides better control and efficiency from their infrastructure so that they can adapt to new cloud-based services. In essence, CloudLinux crash proofs the hosting infrastructure to give you the solid foundation needed for cloud services.
This is why shared hosting is such a good match for what we do.
About Liam Eagle
Liam Eagle has worked as a contributor to the Web Host Industry Review since its inception in 2000, and as editor since 2003. He has been editor of the WHIR's print magazine since its launch. His daily involvement in the gathering and reporting of Web hosting news and his regular interaction with Web hosting leaders gives him an uncommonly broad appreciation of the issues and tends facing the business. Through his WHIR blog, Liam spots Web hosting trends and offers opinions on the industry-wide impacts of major developments and the motivation behind big announcements. Follow him on Twitter @liameagle
No related posts.
