by David Levin, isicad.net
After a 2-year break, in July 2013 LEDAS, not interrupting work on its main specialization - software development services for complex CAD systems and applications - returned to creating its own products and technologies. These pilot projects are exercised by a special unit - LEDAS Labs.
As we approach LEDAS 15th anniversary in April 2014, we asked Alexey Ershov, LEDAS CEO, to tell us about these new projects and how the work of LEDAS Labs correlates with LEDAS efforts, and to describe company’s contacts with Russian partners.
How are things with geometry comparison technology, announced earlier by LEDAS Labs?
An announcement about our project LEDAS Geometry Comparison (this is a working title, in short - LGC) was met with a great interest, which was clearly manifested in requests and established contacts with a number of reputable companies and well-known specialists.
The project has several technological and business components, so at this stage we are moving forward in many directions at once.
First, we are improving the algorithmics and are continuously refining our technology. In particular, we received from a well-known American company a valuable set of industrial models which serves as a basis for algorithm improvements. Also we are expanding the list of supported data formats and increasing utility of the data output for the users.
Second, we are hard at work on implementing a parallel version of our technology. Geometric model comparison is a pretty non-trivial task in the field of geometric modeling, so to achieve the best results it is necessary to explore every avenue that the modern hardware opens.
Third, we agreed to integrate our comparison technology with a cloud PDM-system produced by DEXMA and are actively working together with DEXMA team. On the one hand we announced from the very beginning that LGC project is oriented to cloud use. On the other, PDM is one of the natural areas for applying our new solution. Therefore, cooperation with DEXMA enables us to elaborate side by side on industrial versions in two areas of project development.
Forth, we agreed with C3D Labs to integrate our technology with C3D kernel. LEDAS is a reseller of this kernel so we already have had a possibility to see that this is a high-quality product, well-tested by several generations of KOMPAS users. At the same time, integration with LGC will give an example of using C3D in the tasks of another class, not typical for KOMPAS, thus increasing the status of C3D as a general purpose geometric kernel.
Talking about C3D, one could not but remember an ambitious project on Russian RGK kernel, in development of which LEDAS took part. So why C3D rather than RGK?
We analyzed in depth possibilities of integrating our geometry comparison technology with RGK, and this integration was very efficient in terms of both the architecture and the algorithms. In fact, we completed this analysis and planned integration much earlier than we started moving towards C3D. This integration actually was one of the methods of testing the correctness of RGK architecture and completeness of its set of functions. As I already said, the results of this proof-of-concept stage were truly promising, but moving in this direction was hindered by absence of RGK licensing scheme, which is not yet framed out by the Russian government.
Also the C3D Labs policy with regard to developers of C3D-based products is very attractive, especially when a product is not yet commercialized. Foreign vendors of geometric kernels stick to more stiff schemes of cooperation.
DEXMA is one of the pioneers of cloud PDM systems. Did it affect your choice of the partner? Do you see any additional advantages from integrating two cloud services?
Certainly. For instance, the efforts on integrating DEXMA with Fidesys show that DEXMA ideology allows integrating this system with numerous web-services. You would agree that CAE/FEA is not the first that comes into one’s mind when thinking about expanding capabilities of a PDM system.
In my opinion, comparing geometric models is fundamental function for PDM systems; or, in any case it is an important market advantage. The basic concepts of product data management, such as revisions, are closely connected to the task of geometric data comparison.
A cloud trend in CAD and engineering software is moving slower than in other software areas but it is not disappearing. It seems to me that the speed of its expansion is reduced due to complexity of our industry and the volume of already written code.
Mentioning a mysterious American firm that showed an interest in LGC is intriguing. Can you share any detail?
This American company attracted an increased attention to itself, particularly, this year. The company held a series of consultations with us at the top level, involving some persons that forever made their names a part of CAD history. In spite of it, we looked into technological and algorithmic issues in detail; together we tested the technology and discussed its various applications, including some truly innovative ones.
Parallel computations are one of the global software trends. How far have you advanced in this area in the LGC project?
We advanced in both multithread execution on a single computer as well as in cluster work of a many units transmitting data between the main and additional processes.
Initially we developed a multithread version of our code and the results proved high efficiency of parallelization. For medium- and high-complexity models (with at least hundreds topological elements) experimentally calculated acceleration coefficient for the standard quad core systems was close to the theoretical maximum. The execution time for a code that requires sequential computation constitutes single percents and decreases to below one percent with increasing the model size.
Comparing productivity of our algorithms with the most popular analogues, such as comparison bodies in SolidWorks, we are ahead of our competitors by dozens of times and more, even without clusters, just using a single computer.
Now we are developing a distributed version of our technology, which will be workable on clusters and cloud capacities. In spite of additional overheads associated with transmitting data between computational nodes within a distributed system, the acceleration coefficient is only insignificantly lower than for multithread computations with common memory. Our algorithms are well-scaled: there is no effect of disappeared productivity growth when new nodes are added, which is typical for many algorithms in computational geometry.
It is known that the issue of parallelism in general is one of LEDAS strong points, and a “parallel modeling kernel” project of LEDAS Labs has already been mentioned by media.
Yes, we also research acceleration of the base algorithmic of geometric kernels, Nikolay Snytnikov recently wrote a paper about it. Nikolay, who leads the LEDAS part of RGK project and who is an expert on parallel computations with more than 10-year experience (and, in particular, defended a PhD thesis on this issue), is a key contributor to our research.
Very promising models and new approaches are being built in voxel modeling, which Nikolay characterizes in his paper as a revolutionary way of developing 3D kernels. What I would like to say is that we have made a couple of steps along the path leading from naïve voxel modeling of 3D data through more comprehensive octree to the future of voxel modeling.
Recently we got positive feedback on this paper from representatives of some reputable global brands, whose requirements to geometric kernels are not satisfied to a considerable extent by all existing kernels. Once again we were able to ascertain indirectly but conclusively that RGK performance is as good as the leaders of the global market and has higher performance growth on multi core systems.
How do you delineate LEDAS and LEDAS Labs projects? LEDAS Labs is probably addresses mostly those projects that involve a lot of research?
LEDAS Labs projects are always highly research-focused, but it does not mean that LEDAS is not involved in research. A recent press-release on joint development of geometry search technology with ASCON lifts the veil of this truly ambitious project, where the research part comprises no less than half of all works.
All our service projects are executed by LEDAS and fully meet the highest standards of managing industrial projects, established in our company as far back as during our large-scale cooperation with Dassault Systemes.
You mentioned already a third important example of working with ASCON Group. Does it mean that the companies are forming some special prospective relationships?
To make the picture complete, let’s also remember integrating direct modeling capabilities to KOMPAS 3D, earlier performed by LEDAS. We are happy to work on a broad basis together with ASCON – a leading Russian CAD company that over the last years has paid much attention to implementing state-of-the art solutions. We hope that our cooperation will continue expanding.
Overall, it is nice that recently capabilities of LEDAS as a service company specialized in high-tech development and consulting are sought-after not only by foreign but also domestic companies. Active LEDAS involvement in RGK projects strengthened our long-established contacts with Top Systems, the creator of one of the most technically advanced Russian CADs, is yet another example of not nearly exhausted cooperation. No doubts, such cooperation can be established with other Russian companies. At the moment, it seems to me that Russian companies heavily underestimate possibilities of accelerating their own development by using high-quality development services.
Discussing geometric comparison project, you use the term “Technology” more often that “Product”. Does it mean that LEDAS is not going to create its own products based on this technology?
We position LEDAS as a technology company, which focuses its efforts on innovations, sophisticated new algorithms and architectures. Although previously our main product - LGS was licensed by over than a dozen of engineering software vendors, including such market leaders as ASCON and Cimatron, in general LGS experience taught us that we know how to develop software better than how to sell it. That is why we sold LGS to Bricsys, which has a greater marketing potential.
Now we develop our new ideas such as LEDAS Geometry Comparison to the level of “off-the-shelf” technologies and then we can offer them on the market as end-user products or components. We will be happy to find a software vendor that can become our strategic partner to push technologies developed by LEDAS Labs on the market, as we see our mission precisely in developing smart technologies that can be easily embedded in new or existing products.
[This interview first appeared at http://isicad.net/articles.php?article_num=16648 under the title "Geometry Comparison Technology, Parallel Computations, Cloud Systems and Much More."]