ESPN Case study report
How many requests from users does ESPN receive each second? Do you think this is a consistent pattern or does it have peaks and valleys? When does it peak?
ESPN needs to handle over 10.000 requests per second over a timespan of hours during events. So this is not a consistent pattern. When they have events then they see that there are peaks in user requests. So the valleys will be when there are no events on ESPN.
Why does ESPN store personal information and preferences on its databases and how does this personal information complicate the ability of ESPN to respond to requests from users? Why can’t ESPN just use Web page caching to handle the loads?
They don’t just want personalization, but they also want single sign on. They won’t to have your preference carry forward to your IPhone, other devices or on any other websites that they own.
This personal information complicates the ability of ESPN to respond to requests from users because ESPN has to answer to a user in one millisecond response time. But here they don’t only have to answer to one user, but they have to answer to thousand of them all at the same time. Besides that it also has to be able to “talk” to the correct service and response back to the user in that millisecond of response time.
ESPN can’t just use Web page caching because, they don’t talk about 1 or 2 gigabytes, they need to store over 10 million users. So with full redundancy they need to handle or store 200 gigabyte of data. On top of that they also need to have the same personalization system for their mobile websites, soccernet.com and for deportes.com.
How much information on users does ESPN store? Why does this pose a challenge for ESPN? Can’t they just use a standard 1 terabyte hard drive from a PC? Why can’t they use a single PC?
ESPN stores roughly 200 gigabytes of information of its users. This poses a challenge for ESPN because that have to store so much information that this won’t fit in one JVM, 2 or even 100 of them.
So why can’t they use a standard 1 terabyte hard drive from a pc or use a single pc? Well because of the fact that they have to encrypt the information. They will have to use a program that can transform this encrypted information into something that the computer understands in order to have a fast response.
What platforms do ESPN customers use when access their Web sites, and how does this further complicate ESPN’s processing picture?
ESPN customers use beside the pc also mobile platforms and soccernet.com and deportes.com to visit ESPN. So for all these platforms ESPN has to use the same personalization system. They want their personalized program to be used in all their systems and platforms. So because of the fact that they want their system to work on different devices and platforms they will have to adapt the system to the different platforms and use more data space to keep al the information accurate.
What are the key components in ESPN’s solution? Describe the function of each.
ESPN makes use of a relational database that is called the personalization DB. This database consists of 2 specific components called the Grid and the Composer.
The grid is using ABM WebSphere eXtreme Scale. It’s effectively and an in memory representation of everything in that personalization DB. The key to this is that it is extremely fast, we are talking about a sub millisecond response time. The next component is the Composer. It facades the complexity in dealing with their existing services and it knows how to speak with the grid.
So the composer receives the requests from the users and talks to the grid. So the grid can get your preferences and give that information to the composer. Which will give the information back to you, the user.
Why is scalability so important to ESPN?
When they scale their services, they can give a fast response to each of it users. So how do they do that? Well the grid is based on 10 physical servers. Each of them has a 32-gigabyte RAM. You can dynamically at run-time spin servers down and you can spin servers up when you need them. So they can adjust the amount of servers at the amount of requests.
ESPN needs to handle over 10.000 requests per second over a timespan of hours during events. So this is not a consistent pattern. When they have events then they see that there are peaks in user requests. So the valleys will be when there are no events on ESPN.
Why does ESPN store personal information and preferences on its databases and how does this personal information complicate the ability of ESPN to respond to requests from users? Why can’t ESPN just use Web page caching to handle the loads?
They don’t just want personalization, but they also want single sign on. They won’t to have your preference carry forward to your IPhone, other devices or on any other websites that they own.
This personal information complicates the ability of ESPN to respond to requests from users because ESPN has to answer to a user in one millisecond response time. But here they don’t only have to answer to one user, but they have to answer to thousand of them all at the same time. Besides that it also has to be able to “talk” to the correct service and response back to the user in that millisecond of response time.
ESPN can’t just use Web page caching because, they don’t talk about 1 or 2 gigabytes, they need to store over 10 million users. So with full redundancy they need to handle or store 200 gigabyte of data. On top of that they also need to have the same personalization system for their mobile websites, soccernet.com and for deportes.com.
How much information on users does ESPN store? Why does this pose a challenge for ESPN? Can’t they just use a standard 1 terabyte hard drive from a PC? Why can’t they use a single PC?
ESPN stores roughly 200 gigabytes of information of its users. This poses a challenge for ESPN because that have to store so much information that this won’t fit in one JVM, 2 or even 100 of them.
So why can’t they use a standard 1 terabyte hard drive from a pc or use a single pc? Well because of the fact that they have to encrypt the information. They will have to use a program that can transform this encrypted information into something that the computer understands in order to have a fast response.
What platforms do ESPN customers use when access their Web sites, and how does this further complicate ESPN’s processing picture?
ESPN customers use beside the pc also mobile platforms and soccernet.com and deportes.com to visit ESPN. So for all these platforms ESPN has to use the same personalization system. They want their personalized program to be used in all their systems and platforms. So because of the fact that they want their system to work on different devices and platforms they will have to adapt the system to the different platforms and use more data space to keep al the information accurate.
What are the key components in ESPN’s solution? Describe the function of each.
ESPN makes use of a relational database that is called the personalization DB. This database consists of 2 specific components called the Grid and the Composer.
The grid is using ABM WebSphere eXtreme Scale. It’s effectively and an in memory representation of everything in that personalization DB. The key to this is that it is extremely fast, we are talking about a sub millisecond response time. The next component is the Composer. It facades the complexity in dealing with their existing services and it knows how to speak with the grid.
So the composer receives the requests from the users and talks to the grid. So the grid can get your preferences and give that information to the composer. Which will give the information back to you, the user.
Why is scalability so important to ESPN?
When they scale their services, they can give a fast response to each of it users. So how do they do that? Well the grid is based on 10 physical servers. Each of them has a 32-gigabyte RAM. You can dynamically at run-time spin servers down and you can spin servers up when you need them. So they can adjust the amount of servers at the amount of requests.