Content for the HPC community and innovation enthusiasts: tutorials, news and press releases for users, engineers and administrators
- All News
- Aerospace & Defence
- Artificial Intelligence
- Cloud Platform
- E4 Various
- European Projects
- Kubernetes Cluster
- Latest news
Cloud Computing or traditional infrastructure?
In this article, we can find an interview to E4’s Head of Infrastructure, Ettore Simone, who will lead us in the world of Cloud Computing.
Ettore Simone: everything comes from the will of simplification of a complex and complicated object such as the data center.
Over the years, there has been a trend, i.e. companies equipped with an infrastructure, end up managing it in their own way, without following structured procedures. This has led to having different types of IT departments. Especially in those companies where IT is not the primary function, IT depts have a very low importance.
Over time, however, corporate infrastructures have become very complex, offering a multitude of services. Therefore, there was the need to make an evolutionary leap, as happened for operating systems: to find a solution able to coherently govern the entire infrastructure, from storage to networking, to security, to applications, to virtualization.
One of the infrastructure management solutions is OpenStack. E4, among its high-performance solutions, in fact, offers Fluctus, a cloud platform, based on the most recent and stable OpenStack version.
OpenStack is open source software, so…Why should a company turn to E4 to install it on its infrastructure and pay for a subscription, when the software is completely free? And why choose E4 and not a multinational vendor?
Ettore Simone: Multinational companies that offer their custom version of OpenStack are very often software houses, which therefore have the goal of selling a specific software, a single product, based on OpenStack. Instead, E4 provides a solution that has OpenStack internally, but is created specifically to solve the specific needs of the customer’s infrastructure. The complexity of having to integrate such a large solution as OpenStack into a structured company is very high and requires very vertical skills. E4 has the task of acting as a bridge between a complex data center (deriving from the customer and his specific needs) and the infrastructure solution, also complex (OpenStack), so that the latter can be used immediately and already studied on the needs of the customer. E4 also provides many services, from consultancy, to implementation, to the resolution of post-sales issues.
So why not get the free one?
Ettore Simone: Downloading the entire OpenStack and making it work correctly on an infrastructure is a job that requires vertical skills that can be developed over years: for the individual customer it is not a feasible investment. Moreover, if you start from scratch, there are no guarantees of operation and integration at the end of the process. Finally, if you run into problems, you have no one to turn to but the community.
Above there was mention of a total management level and simplification similar to that of operating systems… So, how was it done before operating systems existed? What are the points in common and contact between a solution for complex infrastructures and the operating system of a single machine?
Ettore Simone: A leap in time should be done! With the introduction of mainframe computers, every component was by itself: storage, computing system, other devices… and all these pieces had to be managed separately, as it happens in traditional infrastructures, both physical and virtualized, where the system engineer must keep track of everything (security systems, hypervisors, etc.).
This is in fact a very uncomfortable way of handling things. The innovation took place with the birth of more complex operating systems, which acted as a filter between users, systems engineer and computer resources and therefore the orchestrators they managed together with storage, memory, network access, services etc… This led to a huge simplification of computer management and optimized the work of the systems engineer, orienting him more to business than to “what is inside the computer “.
OpenStack is doing the same thing but on the entire data center, on a much larger scale, that is to create a layer of orchestration and abstraction between the components underlying the data center (including distributed data centers) and who must use it to provide business services.
Above you said that every company manages its own data center in its own way. It was about the applications or even at a lower level? An infrastructure, with one of the most popular hypervisors, works very well… So, where is the complexity mentioned above?
Ettore Simone: In reality, the hypervisor you are referring to works, and it is true, but it is only a tool that gives you a virtualization service and network management. Other problems of IT management remain. For example, network management, security, storage and applications. How to connect all these parts together? When to renew the hardware and how to manage it? How to avoid disruptions during a hardware replacement? How to make a capacity plan?
Hence, you cannot solve the coordination of everything inside a data center with a simple service for virtualization.
But the vendor producing this hypervisor has many of these functions available with its products…
Ettore Simone: Yes, the vendor gives you some tools, but it does not explain how to structure yourself… and it only works in those areas in which there are its own products. It does not allow you to manage external things, such as network management… In large infrastructures, for example when there was the need to provide services like the cloud, it was not enough to rely on a single vendor, because there is the lack of heterogeneity.
Nowadays, if you compare the control panel of a hypervisor with the panel of a cloud infrastructure, you immediately notice that the first is a technical panel, the second is very simplified, so that users can be autonomous in managing it and also in measuring it and measuring its value.
This is what really differentiates cloud infrastructures compared to traditional ones. As a user, I don’t want to have control of the bit, because I don’t care to have it, I want to have a simplified control panel, with few and clear choices to fulfill.
What if a company already has traditional nodes? Should throw it all away?
Ettore Simone: First of all, nothing that works will be discarded, but it will be used until completely decommissioned. Cloud structures are already widespread in most companies and coexist with the more obsolete. It is possible therefore to have a mixed structure and let the orchestrator choose where running systems.
So, can I manage the nodes of the my virtualized farm with an OpenStack (FLUCTUS) installation?
Ettore Simone: Migration must be done with a parallel infrastructure: the existing infrastructure, once “released”, can be integrated under Openstack and the panel expands, making the user see as many regions as the infrastructures he has installed. From that point on, I have a chance to go and insert my virtual machines or my services in environments, which I choose, or let the orchestrator choose. So from a single panel I can manage virtualization environments sold by Seattle, others virtualization environments sold by Palo Alto or the nodes in which the hypervisor it’s a KVM. The existing machines are then migrated to the new infrastructure. Having this diversity of environments there is also the advantage of diversification of technology: I can instantiate machines particularly vital in clusters, but in infrastructures completely separate. So, if an infrastructure collapses, another one keeps on working.
How risky is the cloud?
Ettore Simone: Whether you choose public or private cloud, the risk is very low. In the case of a private company that chooses a private cloud like OpenStack, this is an infrastructure that is installed internally, therefore the guarantee of security and confidentiality is maximum.
With Fluctus you have your own cloud at home. In fact, we must distinguish between cloud services and infrastructure with cloud technology. The former are services provided externally, while cloud technology can be used at home and this gives user the advantage of having the same simplifications as the public cloud at home. That’s what eliminates the complexity of managing my data center infrastructure, but keeping all the resources internal.
How can I install a cloud infrastructure inside a company? It wouldn’t be too expensive and bulky? Doesn’t it include a lot of hardware?
Ettore Simone: First of all, cloud technology does not necessarily mean “many systems”. It simply means a technology that simplifies the use of resources. You can choose on how many systems to install your cloud, because it is proportional to the services that you have to deliver. You need a minimum number of machines, at least 3…
As I said before, IT is unfortunately often seen as a cost, especially for companies that don’t do IT. With cloud solutions such as Fluctus I reduce the cost, but I also get benefits like:
- Increasing the services I am able to provide
- Improving the performance of technicians who use it
- Breaking down the boundaries between internal and external data center
“Difficult go back once tasted! “
Thanks for all the explanations, but there are two other questions:
E4 Fluctus is based on the stable OpenStack version produced by the community. I wonder how much it is ethical from a company like E4 trying to make money on a product created for free by a community?
And why should I choose E4, such a small company? And most importantly, if I need support, who answers me?
We will answer these questions in the next episode! Keep following us! In the meantime, read more about Fluctus!