voldemort-Distributed Key-value Storage System

voldemort-distributed key-http://www.aliyun.com/zixun/aggregation/9541.html ">value Storage System Blog Category: Network summary Hadoop
voldemort-Distributed Key-value Storage System
Read about the experience of several LinkedIn engineers who built TB-grade Key-value systems: Building A Terabyte-scale data cycle at LinkedIn with Hadoop and Project Voldemort. The specific implementation process has a general description, do not parrot.



In fact, many companies are now facing similar problems in this abstract architecture diagram. Using Hadoop as the back-end compute cluster, what is the most appropriate way to store the calculated data if it is pushed backwards to the front? Put it in the DB again, build index is troublesome, put to Memcached such key-value distributed system, after all, only in memory, the data is easy to lose. Voldemort is a good improvement plan.

Some points worth to learn from:

Keys (key) structure design, a bit of skill;
It is useful for architects to be familiar with the hardware structure. The larger the system, the more so.
With good parallelism. There will be more occasions when the law of Amdahl.
There is one more solution for key-value applications. LinkedIn is also developing a case for this application. If the business type is similar, you may wish to pay attention.



Original address: http://dbanotes.net/arch/voldemort_key-value.html