CDNs are intermediary servers that replicate read intensive data to provide better performance when user requests them. A close relative of CDNs is edge computing (e.g. gaming stations) where lots of computation is done directly close to the user.
Types of CDNs 🟨–
Mainly three types of CDNs:
- Highly distributed ones.
- Database based ones.
- Ad-hoc CDNs.
Advantages and disadvantages 🟩
The main reason we use CDNs is to lower the value of latency: we are in fact bringing the data closer to the user. We have much less data in length to be transmitted. Yet we have some disadvantages too:
- They usually are not able to provide much storage
- Difficult to manage and coordinate.
Cache Design
If you want to optimize for computer caches, see Cache Optimization.
The Reuse Distance
The reuse distance of an element in a certain access sequence is the number of middle elements until we get that element again. In classical LRU caches, with this value you can compute the miss rate.
Increase Hit Ratio 🟩
If we increase the cache hit ratio, we will have indeed better performance. There are some standard ways that are known in the academy:
- Increase cache size
- Design better algorithms
Lazy promotion and Quick Demotion 🟥
These are two design principles that have proven to work when designing efficient caches. Lazy promotion retain popular objects with minimal effort
- Improve throughput due to less computation
- Improve efficiency due to more information at eviction
- Reduces the frequency of recency updates, potentially improving performance.
- The idea is to wait to upgrade an element to VIP member in the main queue, and wait to record a certain number of accesses, or access pattern that enable it to move up.
Quick Demotion: Remove unpopular objects fast, such as one-hit-wonders.
Sieve Caching
SIEVE considers both how frequently an item has been accessed (frequency) and how recently it was accessed (recency). However, it adds an interesting layer: it tries to actively identify and remove items that are likely to be accessed only once or a very limited number of times. These are often referred to as “one-hit wonders” or “transient” items.
The following represents the algorithm in short: !Example of the algorithm
Load Balancing
When we have lots of requests, we want to find an optimal way to split these among the available resources.
Global Load Balancing
- Avoid TCP set up delay (latency)
- Use of DNS caching. AKAMAI uses a kind of iterative DNS discovery (see Livello applicazione e socket#Domain Name System). In this DNS type, it makes a request, and it tells you other DNS name system to contact to get the IP. When you go the IP, you can make the HTTP request to the akamai server for the resource. This server contacts the content provider if there is no copy available, else it does it correctly.
Anycast
Advantages • No extra round trips • Route to nearby server Disadvantages • Does not consider network or server load • Different packets may go to different servers • Used only for simple request-response apps
It uses consistent hashing (See Cloud Storage#Key-value stores) to know exactly what server to contact.