Entirely Russian, convergent and ... hyperconvergent. Why did we make the SCALA-R platform

Talking about testing our product - the first Russian convergent platform SCALA-R for compatibility with ERP systems or the Russian EDMS - we did not dwell on the origins of the project in detail. In this post, we decided to collect the whole “kitchen”: to talk about how the IBS Interlab team went from idea to implementation “in hardware”. We will also show in more detail what our complex is made of and “insert our 5 kopecks” into a lively discussion about the struggle between convergent and hyperconverged solutions.

The world of integrated systems

Since the beginning of the 2010s, there has been a massive shift in the minds of corporate IT customers, and instead of buying hardware and software for each individual information system, many began to think strategically: build a universal infrastructure so that its resources could be distributed to various IT systems as needed.


The growing up of virtualization technologies and the spread of cloud ideology played their role. And although at the very dawn of the "cloud" was meant only public , over time, a private deployment model was formulated and implemented. The clouds in the box did not take long to wait - integrated systems, pre-configured conglomerates of servers and network equipment, the release of which has been mastered by several leading IT manufacturers.

Integrated systems have significantly reduced the cost of creating a private cloud, and, most importantly, dramatically increase the predictability of the success of this event, while simultaneously reducing the deployment and migration time from previous solutions. IBSOf course, as a large and advanced system integrator, he was among the first to start implementing such systems. For several years, we have accumulated significant experience, which allowed us to master these solutions quite deeply and widely. It happened that in the process of gaining experience, they stepped on a rake imperceptibly, but all the problems were answered, and sometimes one only had to be surprised at the grace of engineering and software solutions, and often the systematic and persistence of Western developers.

If not us, then who?

The hardware and software solutions of such giants as VMware, Cisco, EMC, NetApp, Oracle, of course, are wonderful, shelled at tens of thousands of customers around the world, but domestic solutions have grown and matured in almost every cloud sphere: virtualization , network storage, orchestration, management and monitoring, and we are able to assemble equipment. “ Is it really impossible to make the same complex entirely on our, Russian decisions? ”- asked our customers, for whom sanction risks are not an empty phrase. “ Is it really impossible to make these complexes more productive and cheaper - using container virtualization and elegant software-defined storage networks?"- asked those who followed the development of the IT market and did not want to pay extra money only for the well-known logo on the cabinet.

So, having heard these questions and "standing on the shoulders of the giants" - having worked a lot and seriously with almost all the integrated systems present on the global market - we got the idea to create a modern integrated system from the Russian "building blocks". Why not - this is the task of the integrator class, combine hardware and software in such a way, choose the optimal configurations so that everything works smoothly and is suitable for a wide range of customers with a wide enough range of tasks, bring together and make friends of different suppliers in order to quickly and at minimal cost “ mountain ”to produce finished products that could be quickly delivered to our customers.


Basic SCALA-R scheme (and any other convergent PAK)

Taxonomy: platform, convergent, hyperconverged

The natural question is: what place in the world of integrated systems can such a solution take? Here, perhaps, it is worth sorting out the taxonomy of this very world of integrated systems. IDC, following the logic of cloud service models, divides the market for integrated systems into platform systems and infrastructure.

Platform systems- providing platform software resources, such as DBMS, application servers, development environments, this class includes data processing and analysis systems from Teradata, issued since the 1980s, Exa-complexes from Oracle (Exadata, Exalogic, Exalytics), not so long ago, complexes appeared with a ready-made set of solutions from the Hadoop ecosystem. All these products are extremely interesting, many of them can be talked about as masterpieces of engineering, but these are usually highly specialized solutions, in each case for a rather limited class of tasks, and, as you may have guessed, today it’s not them (but then we will definitely return to this topic!) The

second segment - infrastructure complexes- “Infrastructure cloud out of the box”, they allow you to build an infrastructure on which to quickly and with sufficient simplicity and flexibility to deploy and flexibly manage all the IT needs of a modern organization - mail servers, infrastructure of remote workstations, application information systems, file storage. Among the infrastructure complexes, two more subsegments are distinguished: convergent and hyperconvergent . The building blocks of the first are divided according to their purpose: some of them are computing nodes, others are storage nodes, such are FlexPod and VBlock. In hyperconverged - each node combines the role of a storage node and a computing node, the most eminent representatives of this sub-segment Nutanix and SimpliVity.

Hyperconverged systems appeared a bit later, and their market share is still incomparably small, for example, according to IDC estimates for 2014, the global supply of infrastructure complexes as a whole approached $ 6 billion, while according to the same agency, hyperconverged systems sold $ 373 million. However, this share is growing: hyper-convergent ones are growing faster than others on the market, more than two and a half times in the last calendar year! Perhaps because of this, Gartner analysts see the hyperconverged systems segment as a completely independent part of the integrated systems market, not combining, as IDCs do, convergent and hyperconverged into one segment.

And although many of the suppliers of hyper-convergent quite reasonably talk about virtually unlimited scalability, it so happened that their main consumer is organizations that are classified as medium and small businesses. There can be many explanations for this, perhaps the presence of a dedicated high-performance data storage network “big customers” consider critical for loaded transactional applications, the main manufacturers of hypercoherent systems are probably relatively small and start-up companies (while large organizations prefer to work with IT-grandees), perhaps, for large customers, maneuver is important for the ratio of computing and storage capacities with which hyperconverged harder.

Hyper Convergence and Convergence in a Single Box

So, figuring out the alignment of positions and trying to understand what kind of complex we should offer Russian customers, we came to the conclusion that it makes no sense to categorically choose only one approach: since hyperconverged solutions often lack the traditional SAN, and classical convergent solutions lack the flexibility of software defined storage network and ease of scalability, why not connect all the features? It was said - done, by joining in a consortium with several Russian manufacturers, in a short time we designed such a solution, and received the modest name "SKALA-R".

Servers, switches, SAN nodes for the complexes are assembled by the Russian company DEPO Computers , they have designed and debugged all the hardware components.

The virtualization and management platform, as well as the software-defined storage system for the complexes, were developed by Parallels. The uniqueness of their solution is that it allows you to simultaneously run both classic hypervisor virtual machines and containers, and manage all this baggage from a single control panel. Software-defined storage Virtuozzo Storage (still better known under the old name - Parallels PStorage) allows you to turn each server into a node of a single hyperconverged storage network. The "shelves" of the classic storage network are charged with software from the St. Petersburg company Raidix. Monitoring of the complex is provided by software solutions of the Russian developers Naumen and IHD Monitoring.

From younger to older: three hundredths, five hundredths, seven hundredths

Our youngest complex, the SKALA-R 300 Series , can be delivered as completely hyper-convergent, without a high-performance storage network. This option is aimed specifically at customers related to the so-called " small and midsize business"(If it is on a global scale, from the point of view of domestic realities, it is right to call them medium-sized organizations). The cost of the complex starts from 7 million rubles. It is equipped with servers with two processor sockets in a classic horizontal design with a height of two mounting units each (the nodes of Nutanix and SimpliVity are also made in the same design). The minimum configuration is 4 server nodes, and there are no upper limits on the number of nodes, so the "youngest" of this series can be considered conditional, associated with the ability to start small.



SCALA-R 500 Seriesit is equipped with computing nodes with a height of one mounting unit, if the servers for the 300th series have 24 places for installing disks, then the design for the “five hundredths” includes only 8. The price for the most modest configuration is with four computing nodes and four SAN shelves - starts from 7 million rubles. Typical tasks for the “younger series” are, in a sense, universal, “broad-based IT services”: VDI , corporate mail services, group work platforms, IT service systems, whereas on the SCALA-R 500 series and more they can systems with a significant transactional and analytical load should be installed.



SCALA-R 700 Series- complexes of powerful four-processor units in buildings with a height of two mounting units. This series usually also comes with an additional hardware and software module - Acronis backup system. Such machines can be entrusted with business-critical loads, for example, in conditions when the database server requires several tens of processor cores “in monolith”, on one node.



Thus, customers have the opportunity to use the classic SAN, which is in demand in many specific cases, while naturally using unified management tools for the entire structure, while preserving the hyperconverged core and all the beneficial properties of this young class of solutions.

Total

But still, the main thing in this complex for the Russians is its domestic origin. Critics will not be slow to draw attention to the fact that the components of "iron" are produced in Southeast Asia, but now there are no other alternatives. It is much more important that the entire hardware is designed, ordered and assembled in Russia, and the most important thing - the software heart of the complex - is entirely created, developed, maintained in Russia. And in this very fundamental matter, there are no compromises and, as we believe, should not be: with virtualization and control platforms of non-Russian origin, even if assembled on the basis of free software, we cannot build a Russian onecomplex, and those organizations that have experienced firsthand sanctions do not even have to hint at this. As of the end of this year, there are no other such fully Russian integrated complexes, and in this sense, SCALA-R is unique.

Although, of course, in the global world of integrated systems, SCALA-R is a new word. By combining hyperconverged and converged infrastructure, container and hypervisor virtualization, the complex creates new meanings in this fast-growing market, and we are confident that we are contributing not only to Russian business, but also to the global mind :-)