Development Services

We solve the most sophisticated tasks

We develop the fastest and reliable mission-critical software in C and C++. Our company contributes to world most sophisticated open-source projects and has a long and reach research track record.

Watch the fragment of the MariaDB Corporation CEO's keynote at MariaDB User Conference'17

C and C++ Software Development

We provide server-side software development from small proof-of-concept prototypes to entire products from scratch.

The developers of our company have 10+ years of experience in C and C++ development. We develop the fastest code using C and C++ programming languages and x86-64 and ARM assemblers, including SIMD extensions. For example, we outperform Nginx and OpenSSL for more than 80%.

Besides the speed...

We focus on security and reliability of our code. We have discovered vulnerabilities in 3rd party C/C++ projects and analyze our own code for security flaws during mandatory code reviews. Our QA process includes security (penetration) testing.

We also use Python, Golang, JavaScript and Lua programming languages for infrastructure, web, and testing tasks.

Our solutions work in the Linux bare metal and cloud environments, such as Amazon Web Services and Kubernetes.

Evaluate quality and complexity of problems we solve by our open source projects here.

See our open source projects

Tempesta FW

A Linux kernel application delivery controller (ADC) extensively using lock-free data structures, advanced string algorithms, the modern cryptography, and machine learning for adaptive load balancing. Tempesta FW services up to 1.8M HTTP requests per second on cheapest 4-core x86-64 CPU, which significantly outperforms any other open source Web accelerators, including ones employing user-space TCP/IP stacks.

MariaDB Temporal Tables

An implementation of SQL system versioning for MariaDB 10.3. It was the major feature in the 1.3 release. Watch our talks about the project at MariaDB User Conference’17 and MariaDB User Conference’18.

Testimonials

Clutch.co publishes verified reviews conducted by independent analysts. Read the unbiased reviews about our work.

Tempesta Technologies led a transparent and efficient process, meeting the client’s expectations. Their extensive knowledge in their field and commitment to the project enabled them to make effective, business-oriented decisions quickly.

Alexander Pavlychev

Co-Founder, Kinescope

Tempesta Technologies was able to develop and launch the system successfully. There were a couple of bugs reported by the end-users, but the team was able to react quickly and issue fixes within a couple of days. They used Zoom and Slack for daily and weekly communication.

Vladislav Podolyako

CEO, Email Management Company

Tempesta Technologies provided thorough solutions and recommendations. The team also carried out performance testing. They were great i communication and were constantly in touch.

Director

Cybersecurity Platform

Delivering user-requested features and development work, Tempesta Technologies has become an ongoing partner. Their ability to contribute to an existing code base stands out, as does their prompt communication. Future clients will encounter a partner that excels at remediating issues as they arise.

CEO

Database Software Company

Tempesta Technologies is a team of highly dependable developers. They managed to complete all the tasks that were given to them and delivered superior outputs that impressed not only the client but their customers as well.

Vadim Tkachenko

Co-Founder & CTO, Percona

Tempesta Technologies meets the client’s standards of quality and timeliness, leading to a longer partnership. Their clear communication, professionalism, and timeliness make for a further smooth workflow.

Stephan Ilyin

Co-Founder, Information Technology Company

Although the solution hasn’t been implemented yet, the client is happy with Tempesta Technologies’ work. The process was painless and professional, and the team was receptive to feedback.

CIO

Telecommunications Company

High performance

We’re one of the best in development of highly concurrent data structures, multi-core workloads, text processing and sophisticated algorithms using fundamental mathematics. Our HTTPS accelerator Tempesta FW services up to x3 more HTTP requests per second than Nginx. We can make your product to beat any benchmarks of you competitors.

The several facts about our work:

Our TLS implementation, employing the modern mathematical algorithms, is 40-80% faster than Nginx/OpenSSL and provides up to x4 lower latency
Our HTTP parser is not only 60% to 500% faster than Nginx’s one, but it also provides much more solid security checks
We’ve developed the fastest C++ lock-free work queue
We were the first to publish open source benchmarks for the Transactional Memory in Intel Haswell processors and studied its performance in comparison with traditional locking mechanisms
We helped MariaDB to improve concurrency of the transaction processing

There are no “strange performance results” for us and we can investigate quite “magical” cases. Read the case of our investigation of Nginx high percentile latency on a G-Core Labs busy CDN edge node.

Learn why C and C++ are the best for the high performance

We’re very serious about software performance, scalability, and reliability, so we constantly learn the latest technologies in the areas. C and C++ with some Assembly language still constitute the most powerful tool set. Being used together, C, C++, and Assembly languages allow to develop a high-level and reliable code as well as to reach the highest speeds for the performance critical parts of the code.

The fastest and the most popular open source software is developed in C or C++, e.g. the Linux and FreeBSD operating system kernels are developed in C, MySQL, MongoDB and ClickHouse databases are in C++, PostgreSQL – C, Nginx and HAProxy web accelerators are written in C, Envoy and Apache Traffic Server are in C++. The most heavy loaded websites use C and C++ for their server-side logic. C++ is also the main language for the modern deep learning projects like TensorFlow or PyTorch.

C and C++ are among the most used programming languages, so starting a new project in C or C++ means you choose ones of the most used programming languages and thus can find both enough qualified resources and great open source code, which can be adopted by your project.

However, these doesn’t mean that C++ and C are always the best choice. Sometimes you need some quick PoC, just need a job to be done ASAP, or developers of your company work with other technologies. We understand the business needs and do our best to pick the most natural technologies for your business needs. While our company shines in C/C++ development, we also have senior developers in Python, Lua, JavaScript and Go programming languages.

Networking

We’ve been developing high performance network traffic processing solutions since 2008. Our networking solutions are implemented in the Linux kernel, employ kernel bypass techniques (e.g. DPDK and Netmap), or use traditional event-driven sockets in user space. The areas of our expertise are: HTTPS servers and load balancers, clustered DNS services, cluster management software, database synchronous replication, network security (e.g. VPN, WAF, NIPS), DPI, 5G, SDN, and many others.

As an example of our network traffic processing development

We would suggest you to take a look at Tempesta FW – world fastest HTTPS accelerator, which is developed by our company. To make it the fastest one and feature full, we constantly keep our eye on the most popular HTTPS servers and proxies, such as Nginx, Apache HTTPD, Apache Traffic Server, HAProxy, Varnish, Envoy, and H20 (see details in our talk and slides from the SCALE 18x conference). We know internals of the servers and can make any source code modifications or modules extensions for them.

Our team regularly talks at the Netdev conference, one of the most valued conferences about the Linux networking. You can check our talks from the list of our research papers.

Databases

We’re contributors to MariaDB and Percona XtraDB Cluster. We contributed to the most sophisticated parts of the databases: InnoDB transaction engine, Galera Cluster synchronous replication, SQL query optimizer. Besides contributing to the MySQL forks, we delivered projects modifying the PostgreSQL, MongoDB, ClickHouse, and Redis database internals. We also develop ad-hoc distributed databases for network traffic aggregation and analyzing. The typical cases for the ad-hoc solutions are real-time network data mining, web content caching or distribution of machine learning models.

To learn more about our several database projects, watch our talks:

Linux Kernel Extension for Databases, Percona Live 2016, about our open-source database TempestaDB
MariaDB Temporal Querying with AsOf, MariaDB User Conference’17
Querying Data at a Previous Point in Time, MariaDB User Conference’18

Vadim Tkachenko, Co-Founder & CTO of Percona, says about us:

They have unique skills in high-performance system programming and can solve sophisticated technical tasks quickly.

Read this and other verified reviews about our work at our Clutch profile.

Linux kernel

Our company has deep knowledge of the Linux kernel internals. We have experience with the Linux TCP/IP stack, high-speed network adapter drivers (Intel, Mellanox/NVIDIA), XDP and eBPF programming, virtual memory management (VMM), disk and network filesystems(BtrFS, XFS, EXT4, NFS), KVM and LXC virtualization, cryptography API, Linux security modules (LSM).

You can observe Tempesta FW as an example of our Linux kernel projects.

This is more than 100,000 lines of high-performance code with advanced lock-free data structures, modern cryptography algorithms implemented in x86-64 assembler, fast SIMD strings processing algorithms, light weight machine learning, and deep integration with the Linux networking stack.

We contributed into the CloudLinux kernel.

Machine learning

We employ ready-to-use deep learning platforms, like PyTorch or TensorFlow, as well as develop ad-hoc high-performance classification and regression algorithms.

Our implementations of

light-weight clusterization and/or probabilistic algorithms running on CPU SIMD extensions or GPUs can classify millions of objects per second (the typical case for DDoS mitigation).

Read about the case for Email management company. Our team developed the machine learning driven automation system for email management.

How we work

We analyze performance and evaluate security of the current system or design the architecture of the target system.

To provide the best quality and maintainability of our code we use mandatory code reviews, strict coding style, continuous integration (CI) and delivery (CD) systems. See as an example our open source CI for Tempesta FW (developed in Python).

Our C/C++ development team is fully distributed, which allows us to hire the best developers around the world and communicate with our clients almost 24 hours a day.

Why us?

Only proven developers are assigned to your project. Tempesta’s team consists of C/C++ developers, who have been working together for ages. Any engineer working on your project has a proven record on our internal projects. All our developers contribute to open source projects, so you can always observe the quality of our code.

We’re a small team of C/C++ experts passionate about technologies. Even our CEO writes code. Thus, you pay for the high quality code only, not for selling services to you and not for the senior management.

Everything is transparent for you. From the first days we commit code to your repository, so you can evaluate our performance from the first week of the work. You can chat with our developers directly in instant messengers, e-mail, and/or by phone. Our CEO is always on-line for you and is ready to answer any questions.

Case Studies

10 October 2022

Masterhost

We significantly extended features and performance of the Masterhost's massive shared hosting infrastructure.

A FreeBSD kernel module and an Apache HTTPD patch for high performance user switching to execute PHP scripts. The result of the project allowed the web hosting company to switch users much faster than any other existing mechanisms.
MySQL accounting and users limiting. MySQL and FreeBSD kernels were patched. The FreeBSD patch for per-thread rusage CPU statistics was merged into the mainstream FreeBSD source code.
Research in application layer DDoS mitigation algorithms based on analyzing of web users transitions among web site pages (behavior analyzing). The proposed scheme allowed to analyze million of users online on an entry level hardware.

10 October 2022

MariaDB

We have developed the System-Versioned Tables for MariaDB 10.3. The development was commissioned by MariaDB Corporation. We designed the architecture and were responsible for the whole C++ development process, from early prototype to late bug reports from external users. During the project we significantly modified internals of the SQL layer of the MariaDB database, InnoDB storage engine and the Galera Cluster synchronous replication engine. Read an example or our analytics of highly concurrent data structures.

10 October 2022

Percona

Since 2015 we delivered several C++ projects in the database area by request of Percona, including:

Galera Cluster integration into the first version of Percona XtraDB Cluster
Multiple bug fixes in InnoDB storage engine
Porting patches of a custom high-performance MySQL fork from 5.6 to 8.0

Read the review from Percona's CTO and co-founder about our work.

10 October 2022

Wallarm

Wallarm commissioned us to develop libdetection library, an injection and commanding attacks detection library. The library is the part of a NG-WAF, Wallarm. libdetection provides much lower rate of false positives and false negatives than libinjection. Read the review from Wallarm's COO and co-founder about their experience working with us.

10 October 2022

Beeline

Back in 2010, when 10Gbps was quite a speed, we designed and implemented a 10Gbps HTTP traffic (MPLS) analyzer and TOP statistics counter by request of Beeline, an enterprise ISP. 10Gbps performance was shown on an entry-level server thanks to the new technology of data structures construction based on x86-64 virtual memory paging. The core of the project was the heavily modified Linux kernel.
We developed a Hilbert R-tree-based spatial database for real-time analyzing of the TOP statistic collected by the traffic analyzer. C++ was the main language for the project.

10 October 2022

S-Terra

S-Terra engaged us to analyze and optimize the performance of a VPN network capturer (Linux kernel module) to handle multi-gigabit traffic on large x86-64 NUMA systems. To improve the performance of the VPN we introduced

lock-free data structures
NUMA and MSI-X aware scheduling
zero-copy network packets handling

We've achieved stable performance measures with almost 200% performance boost in some of the tests and more than 100% improvement in average. So S-Terra still uses this VPN solution without any slowdowns or client complaints up to this day.

10 October 2022

CloudLinux OS

Our team fixed number of major bugs in the CloudLinux's version of the Linux kernel.