This black-black friday
Habr, we want to remind you that Black Friday discounts have started in Russia. Just for this event, we have prepared for you an article: the history of the appearance of Black Friday and a bit of technical material on how the official online sales service at blackfridaysale.ru copes with peak loads. Look under the cat. And do not forget to take advantage of discounts!
In early May, in the article “Microsoft Azure Cloud in Real Projects” we talked about some companies that have chosen Azure as their cloud service. One of these companies is Black Friday.
')
For the first time the phrase "Black Friday" was applied to the financial crisis: September 24, 1869, the fall of the gold market in the United States. Then two notorious financiers tried to buy up all the gold in the states, hoping to raise the price for it and later make a big profit.
The most well-known version of the emergence of today's meaning of Black Friday is a certain play on words and phraseological units. It is assumed that after a year of unprofitable work (“in the red” - in debt, approx. Ed.), Stores will get a profit (“went into the black” - will become financially successful, ed.) A day after Thanksgiving, because people will spend a lot of money buying Christmas presents.
There is another less well-known version related to discounts on slaves after Thanksgiving in still slave-owning America.
However, in fact, the history of the meaning of the term is connected to the 1950s Philadelphia. Every year, the day after Thanksgiving, the city was crowded with people, mostly tourists, gathered in anticipation of a big football game. Theft of shops thrived in the city, which increased the police headache. It was only in the late 1980s that retail stores found a way to turn black Friday into something positive — what is today called Black Friday.
If you have your own theory - share in the comments.
In this section we want to share with you a short interview with Alexander Usunov.
What loads does BlackFridays experience at times of peak site traffic? How much do these values ​​differ from the standard ones?
On ordinary days, throughput jumps from 0.5-3Mbps, during lunch hours it can go up to 15-20Mbps. Two or three days before the promotion, the indicators go up to 600Mbps and a few minutes before the start increase to 1.2-1.5Gbps. During the start throughput about 7-8Gbps, a maximum of 14Gbps. After about 4 hours, the value goes down to 3-4Gbps and is within these limits until the end of the action.
What did you use to distribute the uneven load of traffic to the site before you decided to switch to Azure?
We used AWS: OpsWorks, API Gateway, scale groups, Cache cluster, Aurora DBMS and CloudFront.
What exactly caused the transition? The main criterion was to save money, time or something else?
First of all, the difficulty of paying large Amazon bills for legal entities. persons from Russia. We also switched to the new version of the framework, searched for the most appropriate SaaS solutions, and Azure experts gave us good options. And, of course, cost savings.
What problems arose during the transition?
The inability to associate a webapp or cdn with a domain without making appropriate entries in the DNS made it very difficult to make a smooth transition. Initially, everything worked significantly slower, we had to reconsider some solutions, now there are no performance problems.
What peak load did the site withstand after the move? Have there been accessibility issues?
We had a division by regions and groups for Russia, Austria and Germany. Each group accounted for about 7-8Gbps in the first hours after the start. In the first minutes there was a small lag on Application Gateway, which gave SSL traffic, they were not “warmed up” before the start. But globally it did not affect the availability of the resource.
How long does it take for you to add additional servers to Azure and how long would it take on your own servers?
We did not use our own servers; deploying additional virtual machines in Azure takes us a little longer than it did in AWS if we deployed from a template. About 10-15 minutes with software and our applications. If you compare the unfolding of the image, then much more time. But we have reduced the direct use of virtual machines as much as possible at the expense of SaaS solutions, and those places where they remained are not time demanding.
Read more about the move here .
In early May, in the article “Microsoft Azure Cloud in Real Projects” we talked about some companies that have chosen Azure as their cloud service. One of these companies is Black Friday.
')
A bit about the history of black friday
For the first time the phrase "Black Friday" was applied to the financial crisis: September 24, 1869, the fall of the gold market in the United States. Then two notorious financiers tried to buy up all the gold in the states, hoping to raise the price for it and later make a big profit.
The most well-known version of the emergence of today's meaning of Black Friday is a certain play on words and phraseological units. It is assumed that after a year of unprofitable work (“in the red” - in debt, approx. Ed.), Stores will get a profit (“went into the black” - will become financially successful, ed.) A day after Thanksgiving, because people will spend a lot of money buying Christmas presents.
There is another less well-known version related to discounts on slaves after Thanksgiving in still slave-owning America.
However, in fact, the history of the meaning of the term is connected to the 1950s Philadelphia. Every year, the day after Thanksgiving, the city was crowded with people, mostly tourists, gathered in anticipation of a big football game. Theft of shops thrived in the city, which increased the police headache. It was only in the late 1980s that retail stores found a way to turn black Friday into something positive — what is today called Black Friday.
If you have your own theory - share in the comments.
Mini-interview with the IT-director of blackfridaysale.ru
In this section we want to share with you a short interview with Alexander Usunov.
What loads does BlackFridays experience at times of peak site traffic? How much do these values ​​differ from the standard ones?
On ordinary days, throughput jumps from 0.5-3Mbps, during lunch hours it can go up to 15-20Mbps. Two or three days before the promotion, the indicators go up to 600Mbps and a few minutes before the start increase to 1.2-1.5Gbps. During the start throughput about 7-8Gbps, a maximum of 14Gbps. After about 4 hours, the value goes down to 3-4Gbps and is within these limits until the end of the action.
What did you use to distribute the uneven load of traffic to the site before you decided to switch to Azure?
We used AWS: OpsWorks, API Gateway, scale groups, Cache cluster, Aurora DBMS and CloudFront.
What exactly caused the transition? The main criterion was to save money, time or something else?
First of all, the difficulty of paying large Amazon bills for legal entities. persons from Russia. We also switched to the new version of the framework, searched for the most appropriate SaaS solutions, and Azure experts gave us good options. And, of course, cost savings.
What problems arose during the transition?
The inability to associate a webapp or cdn with a domain without making appropriate entries in the DNS made it very difficult to make a smooth transition. Initially, everything worked significantly slower, we had to reconsider some solutions, now there are no performance problems.
What peak load did the site withstand after the move? Have there been accessibility issues?
We had a division by regions and groups for Russia, Austria and Germany. Each group accounted for about 7-8Gbps in the first hours after the start. In the first minutes there was a small lag on Application Gateway, which gave SSL traffic, they were not “warmed up” before the start. But globally it did not affect the availability of the resource.
How long does it take for you to add additional servers to Azure and how long would it take on your own servers?
We did not use our own servers; deploying additional virtual machines in Azure takes us a little longer than it did in AWS if we deployed from a template. About 10-15 minutes with software and our applications. If you compare the unfolding of the image, then much more time. But we have reduced the direct use of virtual machines as much as possible at the expense of SaaS solutions, and those places where they remained are not time demanding.
Read more about the move here .
Source: https://habr.com/ru/post/343040/
All Articles