Search/ads and ride/delivery services. As you may know Yandex has very advanced SDC system which was used(SDC robots) by Grubhub in Arizona University. They have delivery services in London/Paris and taxi services in Insrael and couple European countries.
As for the Russia I found news(in Russian) that McDonalds was huge part of orders in delivery service. Ads system fully broken because of Google suspension. Without concurents average prices will rise exponentially. Small business will not be able to buy ads and will closed in near feature.
> how much does language matter in job satisfaction?
I think it depends. I hate JavaScript and Python. My coworkers don't care about language. They are working on the product, the goal is a fully functional service with a huge DAU. I can't complete product-specific tasks. I had been writing low-level software for networks, Linux kernel patches, eBPF helpers/tracers and now I'm working on hardware virtualization. In my free time, I'm prototyping devices on my FPGAs. It's really interesting. I can't imagine that I will drop my current job and will start cooking or guiding.
There is one more reason. When you write something that works in a loop, e.g. web server, exceptions are a huge potential issue. I was working on a C++ project with a very complex state machine. One day our service started using all 128 CPU cores just for stack unwinding. It was an ordinary exception in 3thrd party library raised on every request in the loop. We have tried to make a lot of optimisations in different unwind libraries. But it was far away from the performance of a simple Optional-like errors container as a return value. Exceptions are really Zero-cost if they occur only once when application should be terminated.
I'm working in a company with 20+ years of the legacy code base. We have a custom search engine fully optimized for our purposes. Decades were spent writing this code. It supports all the newest Linux kernel technologies. The codebase is really good. But every year a new inter tries to show that open source solutions are more reliable and faster. They don't try to write code in 1 hr of course. They spend 10 days on average to show proof of concept that is really fast for benchmarking, but it fails when we try to load testing data. I don't know why people think that nobody checked other solutions before them.
The argument wasn't that ES is wrong choice, but that it will take too much effort to integrate and too long to do so.
Group of managers wanted to choose Oracle based technology for this that would further entrench their team and they argumented they can do it faster and cheaper than the development team.
They have already succeeded getting sizeable part of business logic implemented in PL/SQL supposedly to improve performance.
ES would make all their plans obsolete in a hurry.
Obviously, they were not interested in improving the product but in grabbing power, budget and control of the development process. Also they needed to score some points to show they accomplished something during calendar year.
Of course development team was also slow to deal with problems and overworked with other projects. This only fueled actions of database operations.
So my actions angered both managers of database operations as well as development teams.
> Group of managers wanted to choose Oracle based technology for this that would further entrench their team and they argumented they can do it faster and cheaper than the development team.
That's true. Most Open Source solutions don't work on huge workloads. You can read about Hadoop at Facebook or MySQL. They fully rewrote MySQL engine and partially Hadoop. Oracle solutions work well on huge workloads.
This company is not a software developer. The software department is just another team inside the company. The first goal of any non-software company is to reduce costs for IT infrastructure and development. It's not Google or Facebook where software development is a key factor of company capitalization.
> Obviously, they were not interested in improving the product but in grabbing power, budget and control of the development process. Also they needed to score some points to show they accomplished something during calendar year.
Or just do their job: provides a reliable service for storing data. The main goal of the DB/SysAdmin department is to show 99.99% uptime. The software department could not guarantee this uptime. Of course, if all these software developers are good SRE that's another situation. But I'm sure no one has experience of supporting huge systems for years.
> So my actions angered both managers of database operations as well as development teams.
If you really want to change something it will be better to make a virtual team from both departments and discuss all pitfalls from both teams. Maybe you are solving non-existent problem.
> That's true. Most Open Source solutions don't work on huge workloads. You can read about Hadoop at Facebook or MySQL. They fully rewrote MySQL engine and partially Hadoop. Oracle solutions work well on huge workloads.
Most “huge” workloads are not actually Facebook-huge.
Because of the misinterpretation of Baseline Requirements for stateOrProvinceName field lot of CAs issued certificates with ST = Italy/France/etc. Now Firefox is forcing to reissue thousands of certificates in 5 days.
Yes and no. If you are living near base station in a big city it will work. But in the countryside where no good connection from ISP equipment to the base station the bottleneck will be this link. Furthermore, 5G uses a higher frequency than LTE, it could be not cost-effective to cover remote regions with small number of users. Starlink could solve problems with cables, but adds problems with RTT. Maybe one day all FAANG companies will open data centres on Earth orbit.
> Maybe one day all FAANG companies will open data centres on Earth orbit.
I know you are joking, but the weight, power, and heat dissipation all make this impractical. They tried similar ideas with cargo containers and underwater enclosures, which were a lot more practical than outer space, but they didn't stick with it.
I think the main problem is spare parts. A long time ago I was working in the data centre. I don't remember any day when there was no problems with hardware.
The signal goes through the ionosphere to the satellite than to another satellite which has a link to the Earth. Starlink uses LEO, so there is 2400 km link(1200 km from/to the ground). 1 ms is time for the signal to travel for 300 km in a vacuum. So theoretically the lowest RTT could be about 8 ms. Right now RTT to my ISP is 2 ms.
I'm was born in small-town in Russia with 5k people. There was an area with private houses and 2-3 floors houses with 18-24 apartments. We have linked attics of buildings with apartments with cat 6e cables and cheap D-Link switches. In 2002 our telephone hub was replaced by a digital telephone switch station where we were renting five 2 Mbits ADSL channels. So users were connected to switches in attics for a small price and anyone who wants access to the internet could configure VPN to our FreeBSD "router". I don't remember the speed limitations, but the cost of 1 Mb was about $0.04(2 Russian rubles). In 2004 ADSL was upgraded to ADSL2+ with 20 Mbits per port. In 2006 all copper equipment was replaced by fibre channel cables/routers. And some people from private houses connected to our network(they were using dial-up). In 2010 huge Russian ISP Rostelecom bought our small network. At the end of 2006, I moved to Moscow because I went to college. And of course, me and 3 of my new college friends has built local ISP in university's dormitory :) The new dormitory was opened after two months after the start of the school year, so there was no any network. Officially it was not possible to connect this building to the internet. We rented a Wi-Fi router in someone's apartment in the building next after our dormitory. I don't know what happened with network because I was expelled from university :)
Well, "ssl_early_data" is opt-in. If you enable it on a virtualhost, then you also need to look at the "Early-Data" request header in your backend and make a decision there. e.g. process GET requests, otherwise send HTTP 425 Too Early.
It does seem a bit unsafe. An administrator might opt-in because they copy-pasted it from a tutorial, and not understand or pay attention to the second part.
I think it will be better to fully disable early data for people without full control of DC's network equipment. I don't know why Cloudflare made a decision about using headers and Too Early response. They have full control of their POPs. It will be better to measure RTT and use UDP based KV storage with tickets only for clients with high RTT. So for clients with RTT higher then access to KV storage it will be better to issue tickets, for other clients it will be better to drop early data and use full handshake. Currently, I'm working on a project with the same idea.
To measure RTT you need to perform a round trip. Hence the name. But the _whole point_ of this feature is to avoid incurring the cost of an extra round trip if possible.
There is no need to send extra data to measure RTT. On the TCP handshake SYN/ACK you already know RTT. Linux kernel provides this info in tcp_info data structure.
From Power9 cluster with array of GPUs? For work loads like http servers/php code/MySQL/etc intel CPU will be much faster. We have CPU samples from all vendors, unfortunately Intel right now is the best platform.
That didn't answer my question. That's fine if you think Intel is better (I don't, but that's orthogonal), but that doesn't make PowerPC less general purpose.
As for the Russia I found news(in Russian) that McDonalds was huge part of orders in delivery service. Ads system fully broken because of Google suspension. Without concurents average prices will rise exponentially. Small business will not be able to buy ads and will closed in near feature.