Relational databases give you a lot more layout and query flexibility, and are easier to use, but you also end up with more pain if you later need to partition your data if it won't fit on a single database server/cluster. BigTable and similar things scale essentially infinitely, but do so by not offering a bunch of relational database features and making the design problem a lot harder.
Given probably 0.1% of places end up needing the BigTable level of scalability, you're almost always better off going with a relational database. If your data model is simple enough that either one would work, I'd be inclined to still pick a relational database.
Size, redundancy and performance are all key words that pop up in my head.
It's fairly easy to create a little database that does its job but when you need to do significant scaling and redundancy, while keeping it performing at the speed of light, it's cheaper for lots of people to ship the job off to Amazon who've been doing it for a lot longer than most of us.
发布评论
评论（2）
关系数据库为您提供了更多的布局和查询灵活性，并且更易于使用，但是如果您以后需要对数据进行分区（如果它不适合单个数据库服务器/集群），那么您最终也会遇到更多的痛苦。 BigTable 和类似的东西基本上可以无限扩展，但这样做是通过不提供一堆关系数据库功能并使设计问题变得更加困难。
鉴于可能有 0.1% 的地方最终需要 BigTable 级别的可扩展性，因此使用关系数据库几乎总是更好。 如果您的数据模型足够简单，任何一个都可以工作，我仍然倾向于选择关系数据库。
Relational databases give you a lot more layout and query flexibility, and are easier to use, but you also end up with more pain if you later need to partition your data if it won't fit on a single database server/cluster. BigTable and similar things scale essentially infinitely, but do so by not offering a bunch of relational database features and making the design problem a lot harder.
Given probably 0.1% of places end up needing the BigTable level of scalability, you're almost always better off going with a relational database. If your data model is simple enough that either one would work, I'd be inclined to still pick a relational database.
尺寸、冗余和性能都是我脑海中浮现的关键词。
创建一个完成其工作的小型数据库相当容易，但是当您需要进行显着的扩展和冗余，同时保持其以光速执行时，对于很多人来说，将工作交给亚马逊会更便宜，因为他们已经这样做的时间比我们大多数人都长得多。
Size, redundancy and performance are all key words that pop up in my head.
It's fairly easy to create a little database that does its job but when you need to do significant scaling and redundancy, while keeping it performing at the speed of light, it's cheaper for lots of people to ship the job off to Amazon who've been doing it for a lot longer than most of us.