An open microphone from DevOps Deflope, stories about the infrastructure of Skyeng and Nvidia and not only

Hello, next Tuesday, warm lamp gatherings are planned on Taganka: there will be Artem Naumenko with a story about infrastructure as a product, Vitaly Dobrovolsky with a report on balancing the Kafka cluster and leading profile podcasts with a secret topic for discussion so far. And we are also expecting a special guest from the northern capital - Vitaly Levchenko, the organizer of the St. Petersburg SRE party.

UPD. The seats in the hall are over, but we will broadcast on this channel . Get connected online!


In the meantime, we talked with the speakers about how the life of SRE-specialists in their teams works and what companies need it at all.

Artem Naumenko and Vitaliy Okulov, Infrastructure Managers at Skyeng

- What are SRE doing in your teams?

- We are just starting to attract such people to the staff to ensure the smooth operation of our many projects: billing, CRM and others. Skyeng has a diverse architecture, many servers, a different technological stack and different databases. We want SRE, together with development teams, to plan and implement changes in the infrastructure, automate work, etc.

- What interesting tasks does your team solve?

- The current interesting project is the clustering of the company's public services packaged in Docker. Only an experienced team of professionals can plan and implement such a change in several projects.

- Which companies need SRE, in your opinion?

- Large companies that put the stability of the service in one of the first places, and at the same time grow 2-3 times a year. Like us, for example)

Vitaly Dobrovolsky, Systems Engineer at Nvidia

- What tasks does your team solve?

- Our company is constantly looking for new opportunities for using GPUs in various fields, so we are experimenting with introducing them into our processes. We have a geo-distributed infrastructure with data centers in different locations, so many tasks are solved with an eye to it. Integration projects never have a solution in advance - tasks that require a non-standard approach will always meet on the way.

- What do you do with SRE?

- We are engaged in processing events from various sources for subsequent aggregation and visualization. Our operations provide stability and availability of the input bus (we use Kafka), raw data processing (YARN, Spark), storage layer (HDFS, Hive, ElasticSearch) and visualization layer (Kibana, Tableau, OmniSci Immerse). In addition, we integrate with other teams and their stacks to join efforts in data processing.

- And which companies need SRE, in your opinion?

- The SRE model is strongly tied to the organizational structure and career cycles of specialists. For example, almost all specialists at Google begin their journey with SWE (developers), and when they gain experience, they can form an SRE team from them, which is connected to the products. That is why Google called this DevOps pattern “class SRE implements DevOps”. If the company has the opportunity, and the business structure allows you to form SRE teams and connect them to products at different stages of the life cycle, they can use this model. Otherwise, the company should implement DevOps practices in a different way. In the end, both SRE and DevOps have the same goals - building flexible and resilient systems.

pps We are waiting for everyone who wants and can, at the meeting on August 27 at 19 o’clock , and everyone who wants but can not - on the air with seven with a penny in the evening.


All Articles