|
|
At the moment, in the IT environment, systems that process a large number of requests per unit of time are considered highly loaded - conditionally from 10 thousand operations per second. And to which high requirements are imposed on availability and stability - practically 24/7. The number of users here does not matter.
High-load systems include billing systems of mobile operators that calculate the subscriber's balance and control the time of their dialogues, the banking sector, stock exchanges (with the proviso that there are windows on the stock exchanges when no trades or transactions are conducted). All modern neural networks such as ChatGPT are also examples of high-load systems, because they process a large number of requests in two modes. On the one hand, this is the accumulation of knowledge and constant training of the model, on the other hand, the processing of user requests.
This is why, as many have noticed, the processing of a request in ChatGPT is not instantaneous. The system needs time to accept the request, look through its databases, how to content writing service respond to it, build relationships and give the information out.
WHAT BUSINESS FACED
Some companies were prepared for the current situation and had already switched to import-independent software or developed their own. For example, Yandex created its ClickHouse database management system. Those who stored data in the clouds experienced difficulties, but, as a rule, large corporations store data on their servers or in their private clouds.
Those companies that had been investing in software from global giants Oracle and SAP for years suffered especially badly, as vendors officially limited or stopped technical support, and also reduced the possibility of receiving updates and renewing licenses (we are not taking into account various workarounds).
Therefore, everyone was faced with one and only choice: to switch to import-independent software and change the technology stack before the current licenses expire and the “switch” is turned off, as happened at some enterprises with foreign participation.
Many companies' software problems were supplemented by hardware problems: for example, servers from IBM or Oracle, which left Russia, stopped working, and it was impossible to quickly buy a replacement. Or even if they worked, what to do with them if they are tailored only for products from now unfriendly companies?
Companies were urgently trying to get new servers, and the old ones were most often sold to the CIS countries. An engineering task that is usually solved over several years required a solution in a very short time and under conditions of severe external stress: a shortage of specialists for this type of task and equipment.
The transitions were multi-stage: first they changed the hardware, leaving the software, then they changed the software
Everyone had to go through the difficulties of conversion, because it is simply impossible to smoothly move from one database to another.
Unfortunately, there was some data loss during the transition. Vital information remained, but minor, auxiliary data had to be deleted.
CHANGES IN BUSINESS PROCESSES
Domestic software has a different interface and capabilities from foreign software, but 70% of it is definitely suitable for completing the necessary tasks. Yes, you need to get used to it and forgive some moments, understand that there will be other fonts, colors and buttons, some functions are implemented differently. Often, corporate clients want everything to be as they are used to, but alas, you will have to relearn. And we will not repeat everything that you have already read and heard this year: domestic software has a long way to go, but it is also worth recognizing that the speed set by the situation gives a fantastic pace to development.
|
|
發表於 2024-11-9 12:07:37
收藏