I would agree with @Shiraz about proof of concept, in particular, ensuring the parts of the system communicate appropriately. It is possible to combine elements of his answer, to gain on both the time front and the expertise front (Details below). I would say that I would view such an assessment as you are contemplating as essential - it could well be the difference between a successful project and a time and money sink.
I have been involved in assessing architectures for a system to serve the need of 3000+ software developers who were developing for multiple platforms. This system was not as data intensive as yours, but we had previously done work assessing database capabilities and sizing for the production system, which was about 50% bigger than your data.
One of the steps we chose to do with the database evaluation was to set up a joint project with academia to assess the performance of very large systems. This was when RDMS were newish, and we could find no-one who had experience of the sort of usage we envisaged. Both parties gained from this; we got the results we wanted at a much reduced cost over consultancy or getting in experts; they got a substantial research project, and funding for it.
The way we went about assessing the architecture for the development environment we drew up a requirements specification and then put it out to tender. However, the element that was distinctive about the tendering process was that we requested those on the short list to present a vertical slice of their proposal at a week long fair with all the other developers there on competing stalls. We then provided opportunities for all the developers attend the fair to ask questions, test out the proposals, and give us feedback on their experiences. About 45% of the developers took advantage of the opportunity, and were able to intensively scrutinize the competing solutions. This gave us combined expertise and difference of views that tapped into the internal resources we had available, and prompted the vendors to provide substantial technical input into the process. It was also extremely quick (about 6 weeks setting up this part of the tender, for 3-4 people, then the week of the fair, followed by a couple of weeks compiling and comparing the data arising from the fair.
I mentioned the question of communication between elements of the system. In both the cases mentioned, it was not enough to have assurance that (for example) both parts used SQL; or that both parties used standardised SQL; or that they claimed to adhere to standards; it was necessary for them to demonstrate that SQL generated in one part of the system could be read and manipulated by another element of the system. This provided a few shocks for both us and the vendors - who often were not as adherent to standards as they claimed.
发布评论
评论（2）
需要做的一些事情是：
使用您 要做出此类选择的基于透视的体系结构，请参阅 http://msdn.microsoft。 com/en-us/library/bb245776.aspx
Some things to do would be:
We us Perspective Based Architecture to make these kind of choices, see http://msdn.microsoft.com/en-us/library/bb245776.aspx
我同意@Shiraz 关于概念验证的观点，特别是确保系统各部分适当地通信。可以结合他的答案的元素，以获得时间方面和专业知识方面的优势（详细信息如下）。我想说的是，我认为您正在考虑的这样的评估是必要的——它很可能是一个成功的项目与时间和金钱消耗之间的区别。
我曾参与评估系统架构，以满足 3000 多名针对多个平台进行开发的软件开发人员的需求。该系统不像您的系统那样数据密集，但我们之前已经完成了评估数据库功能和调整生产系统规模的工作，该系统比您的数据大大约 50%。
我们选择进行数据库评估的步骤之一是与学术界建立一个联合项目来评估大型系统的性能。当时 RDMS 还很新鲜，我们找不到任何人有我们设想的那种用法的经验。双方都因此获益；与咨询或聘请专家相比，我们以大大降低的成本获得了我们想要的结果；他们得到了一个实质性的研究项目和资金。
我们评估开发环境架构的方式是起草一份需求规范，然后将其进行招标。然而，招标过程的独特之处在于，我们要求入围名单上的公司在为期一周的展会上与所有其他开发商在竞争摊位上展示其提案的垂直部分。然后，我们为所有参加展会的开发人员提供了提问、测试提案并向我们提供他们的体验反馈的机会。大约 45% 的开发人员利用了这个机会，并能够深入审查竞争解决方案。这为我们提供了综合的专业知识和不同的观点，可以利用我们现有的内部资源，并促使供应商为该流程提供大量的技术投入。速度也非常快（大约需要 6 周时间为 3-4 人设置这部分招标，然后是展会一周，然后是几周的时间来编译和比较展会产生的数据。
我提到了这个问题在上述两种情况下，仅保证（例如）双方均使用标准化 SQL 或他们声称遵守标准是不够的；他们有必要证明系统的一个部分生成的 SQL 可以由系统的另一部分读取和操作，这给我们和供应商带来了一些震惊 - 他们往往并不像他们声称的那样遵守标准。
I would agree with @Shiraz about proof of concept, in particular, ensuring the parts of the system communicate appropriately. It is possible to combine elements of his answer, to gain on both the time front and the expertise front (Details below). I would say that I would view such an assessment as you are contemplating as essential - it could well be the difference between a successful project and a time and money sink.
I have been involved in assessing architectures for a system to serve the need of 3000+ software developers who were developing for multiple platforms. This system was not as data intensive as yours, but we had previously done work assessing database capabilities and sizing for the production system, which was about 50% bigger than your data.
One of the steps we chose to do with the database evaluation was to set up a joint project with academia to assess the performance of very large systems. This was when RDMS were newish, and we could find no-one who had experience of the sort of usage we envisaged. Both parties gained from this; we got the results we wanted at a much reduced cost over consultancy or getting in experts; they got a substantial research project, and funding for it.
The way we went about assessing the architecture for the development environment we drew up a requirements specification and then put it out to tender. However, the element that was distinctive about the tendering process was that we requested those on the short list to present a vertical slice of their proposal at a week long fair with all the other developers there on competing stalls. We then provided opportunities for all the developers attend the fair to ask questions, test out the proposals, and give us feedback on their experiences. About 45% of the developers took advantage of the opportunity, and were able to intensively scrutinize the competing solutions. This gave us combined expertise and difference of views that tapped into the internal resources we had available, and prompted the vendors to provide substantial technical input into the process. It was also extremely quick (about 6 weeks setting up this part of the tender, for 3-4 people, then the week of the fair, followed by a couple of weeks compiling and comparing the data arising from the fair.
I mentioned the question of communication between elements of the system. In both the cases mentioned, it was not enough to have assurance that (for example) both parts used SQL; or that both parties used standardised SQL; or that they claimed to adhere to standards; it was necessary for them to demonstrate that SQL generated in one part of the system could be read and manipulated by another element of the system. This provided a few shocks for both us and the vendors - who often were not as adherent to standards as they claimed.