by:dbanotes@gmail.com(Fenng)
Author:Fenngpublished in thedbanotes.net.
book toabove. Video please visitInfoQ.
InfoQ Chinese stations: in the Web 2.0 era of massive data for the growing number of developers, is no longer a distant topic, which may not visited a lot of Web2. 0 sites are likely to have a Zashe amount of data, then for this site, in addition to optimize the database, in addition to the cache, and then there are those strategies »
Fenng: I think that is possible in the storage area will have some big challenges. For example, the reliability of storage, as in the past have been BSP service providers to customer data does not even backup, led to the loss of many users for some time within the data, so the whole site has great influence on the reputation of the user experience Also very bad.
With the rapid development of Internet, data is generally more expansive, in the process, how effective data storage and access to effective, is a more complicated issue. We said earlier a lot of Web 2.0-related topics, from another angle - for example, [my company-paid, is also faced with such a large volume of data, massive data challenge. At present we have a strategy, but also follow the SOA strategy of thinking is that the database related to certain data services SOA treatment. Another strategy is more importantdata life cyclemanagement, we have this, in the data life cycle has been completed, will do some data related to the handling of the archive, to conduct secondary storage Or tiered storage. Then the other hand, some Web 2.0 site, I think this can also use the idea of mechanism: the user has visited or is unlikely to visit the relatively low frequency (data), a tiered storage, or do something extra to visit the formulation of strategy , Is very necessary.
InfoQ Chinese stations: We have also heard of another sub-database system, then you talk about the film is how this strategy-a strategy »
Fenng:--In general, it is not a relatively new technology, MySQL in 5. x version, a district function. Well before this, MySQL is not the function of zoning. At that time, if the need to address some rather large amount of data, such as to the time on the basis of historical data processing, will be relatively trouble. People may be in line with local conditions, have Sharding this technology strategy.
, strictly speaking, data-points in the past, we actually have some relevant practice in other (type) of the database, we also have some history of strategy, it was not fully defined the term down. As far as I know, the word is from the large-scale online games in development of people. Most users will be concentrated in a particular region. This part of a regional focus on the user, they will be placed on a specific server. Between different regions of the correlation between the user may not, the scenes and we are now a database patch strategy is very, very similar, we present the database to do if some of the films, will also use this basic idea, For example, depending on the scope of user ID, or different areas (to patch).
If business is built on site, according to the type of product could do, we will of different products of different types of data thrown into the DB, such correlation between the DB is very small. In between the DB and then we may have a package layer, layer in this package on the application users, as is transparent, then it reached a high degree of expansion of our database on the goal.
InfoQ Chinese station: the sub-What are the pros and cons of such a strategy? »
Fenng: First, the benefits of the film still very easy to see. At least we can achieve the DB does not depend on a single point, and this can be done smooth expansion, as we often say that the Scale Out (horizontal expansion) mechanism. It is also the more obvious defects, high-speed processing services for the Web site, it has its own deficiencies, in fact many friends should know that if a cross-database services, so the designers, the coding staff, It is quite difficult. So if a transaction even if the two cross-over DB, Sharding complexity will be very high. Sharding industry in the application of this scene is basically re-read comparison of the situation, but also for the safety of Yaoqiubugao services, such scenes will be very suitable.
[last month wrote a chapter Sharding things to the "programmer", does not know when it published out -
InfoQ Chinese stations: in many sites in the design of the structure of the vast majority of projects in the area of lasting relationship is a data mapping (ORM) approach. For all this high load of large-scale web applications, what are you think of it »
Fenng: First of all, I Xiangna-we pay for, ORM we feel very good use. In a relatively large development environment, the development team, it's shortcomings may not easily seen from the large. Because we use the ORM, it is easy to DB middle layer completely isolated from this. Then this level of expertise to deal with the SQL DB staff ---- our side there are special development DBA, by them to focus specifically on the monitoring of the management, planning and even some kind of work. This development engineers are architects here, they can focus on doing more in other areas of investment, a big team like ORM I think these still very easy to see the benefits.
[ORM have a relatively good place to security, SQL injection can be effective in reducing the hidden danger:
on the other hand, we look at its shortcomings, because, like a number of small and medium-sized Web site, also may be relatively less manpower, we use (development) tools (or framework), it might as PHP, ROR these things, That is, in the development, is relatively easy to use. So this time, in fact a potential problem is that when the code size to a certain degree, if not do something ORM, it may give the site some potential for instance, code management problems, this is only my personal View, in fact everyone in the application of specific scenes may have their own headache, I not an expert in this area, only your reference.
InfoQ Chinese station: you pay-what, in fact, the enterprise-level application in the enterprise-level application used by the framework of this strategy and general Web 2.0 site used by the framework of this strategy will be What similarities and differences »
Fenng: In fact, it is clear that the pay-in factbusinessis a very complex [have misunderstood part-payment business is very simple: this and we Many of the Web2.0 companies do not like, Web2.0 it could get from one point intercept. At this point, I think the comparison done thoroughly. Bao is paid, it may be a bit like we had to do some common software, he has to consider different industries, different users, as well as between the sale and so many of the relationship between banks and so on, this complexity is great The.
This is in fact a certain extent on the decisions of our Web 2.0 companies and the application of different solutions, like the current pay-in, a year ago, even two years before they consider our entire site SOA, component-based. In this process, is also considered a number of Web 2.0 as the technical elements, but overall the idea of it, or say to the SOA, service-oriented in this regard to the step entered, and then from the SOA on this point, in fact Many Web 2.0 companies, they may not be able to complete the realization of the complete achieve such a service-oriented, I think this may be one of the two very different surface characteristics.
addition, the pay-as some are trying to do, to external customers, to provide some services interface, or even a completely open platform, which also like these and our current FaceBook, or is that the MySpace such as the United States The social area, SNS network of some common.
InfoQ Chinese points: that in Web 2.0 site inside this area, the site's structure What are the main trend of below-how will also have to be a «
Fenng: In fact, as a technical staff, whenever want to talk about trends, we certainly want to laugh. The medium and long term, some of the domestic new Web 2.0 services gradually emerging, with the development, I believe that more commercial elements and come. Is a lot like before Web 2.0 company is using open-source technologies, with the scale of the expansion, previously provided some of the open-source technology organizations or individuals they will try to carry out some commercial operations. Commercialization is not a bad thing, on the one hand to us to provide better service. On the other hand, they have enough business support, in turn, after they examine the open source development environment, a good development environment for. I believe that in the next two to three years, will be part of the commercial Web 2.0 companies rushed to the development of eco-circle inside.
then in terms of technology, such as MySQL months before the acquisition by Sun, at least in the Web 2.0 software in the chain of such an important part of (MySQL), some people may feel to a number of issues. But now in the database like this one, it does not rule out like these other PostgresSQL database, taking advantage of this opportunity to be embraced by commercial companies, they will make some more large-scale application scenes out. In the database in this area may limit you, several open-source database to form a deadlock, Sun in some
original link:http://www.dbanotes.net/arch/infoq_interview_review_2.html
No comments:
Post a Comment