Sina Cloud Computing senior Technical Manager Conglei: NoSQL application in SAE

Today, the arrival of "big data" is no doubt, especially in telecommunications, finance and other industries, almost to the "data is the business itself". This trend has made a lot of changes to companies that believe in the power of data. At this time, in order to allow more people to understand and use the analysis of large data, CSDN's exclusive large data technology conference was held today in Beijing CTS building. The conference brings together Hadoop, NoSQL, data analysis and mining, data warehousing, business intelligence and open source cloud computing architecture and many other hot topics. Including Baidu, Taobao, Sina and other industry renowned experts and participants gathered together to explore the big data wave of industry response law and the big data era of choice.

Conglei, senior technical manager of Sina Cloud computing

Sina senior technology manager Conglei said the 2011 Sina SAE platform registered users have reached 50000, the application of more than 100000, daily PV reached 100 million, active developers reached 10,000.

Conglei also introduced Sina's own development of KVDB,KVDB to support the public cloud computing platform for the massive key-value storage. KV DB Support storage capacity is very large, 100G storage space for each user, can support 1 billion records, users can use KV DB storage simple data, such as friend relationship. KVDB has the advantages of storage engine substitution, arbitrary module level expansion, support for read/write separation, support for prefix lookup, support for secondary index, support for authentication, support for heavy balance and seamless migration.

The following is a transcript

Hello everyone, I am glad to share with you here a topic about SAE in NoSQL. If you have some ideas about SAE, and comments, you can also pay attention to Sina official Weibo. In addition, Saejava platform, has been in the beta, we are interested in the official micro-blog to apply for testing channels, to join us to test, we together to improve the SAE. Today, a brief report on the SAE development, this picture is the SAE development of one, relative to a milestone, from August 09 SAE Cloud Computing Group was established, was very small only a few people, November 09 SAE released a version, to this year exactly 2 years, By 2010 SAE released a lightweight cloud storage product micro-disk. This May also has a lot of open registration, now anyone to use the SAE do not need any invitation code, approval process, as long as the Sina account can be used.

Now the SAE opened the payment, SAE also classified as Sina cloud computing, there are some third-party sites, the Internet Advisory Class site also ran to the SAE. Then, in the SAE products are mainly computing services, storage class services, there is a cloud application store and cloud services store CDN. About the Cloud App Store and the cloud services store, which we've heard about, like the App Store, but we know that App Store is either based on Apple iOS or Android, and SAE if it's not OS, our OS is the Internet, app on the Internet. Store, you now in the SAE only need to spend 30 seconds to open a own group purchase site, you can open a forum, photo albums, Wikipedia sites, to do the Internet app Store.

What is the service store, in turn? We as a developer, the things you develop do not necessarily have interfaces, some people develop things, such as I am Apple language developers, I developed this thing is very valuable but there is no interface, this thing you developers are trying to sell his API to users, this time can actually use the SAE Division store, It is a service concept to carry out the whole statistic, log, and report process, and you sell your API architecture on top of it.

Look at now SAE development of three indicators, one is registered users, the current SAE registered users are mostly developers, although the number is not much, but the quality is very high. In particular, the SAE to do the developer certification, if you use the SAE words should have heard, any one as long as the developer's serious can get a considerable amount of cloud, the equivalent of the SAE to the real developer free money let him in the SAE development application. Another application number, the number of applications is currently 100,000, daily PV more than 100 million, there should be several billion.

We also looked at the SAE running on these applications and services are reliable and unreliable? This is a Q3 for 45 minutes, downtime 4 times, total time 56.05. Looking at more than 10,000 active developers, just mentioned the developer certification, the SAE is actually more focused on the ability to create value core developers, which mainly refers to external developers, including the mobile Internet domain. Of course, the SAE and the official cooperation of PHP, if everyone is a lover login php, the current PHP is the only official website in the mainland is the SAE provided, which shows that cooperation between the two is also strengthening, this piece of our cooperation with the official will also be strengthened.

The last one is the application store, what are the applications, this piece is a list, not to say, Weibo,hdwik, group purchase and so on. Starting from this page today on the topic of technology, our topic today is MySQL on HCE, I speak SQL today, I personally graduated from 06, 07 began to do cloud computing development. At that time we were looking at Amazon grew up a group of people, Amazon think that SQL is not important, this is the Amazon cloud computing, because he felt he could launch his own products, this product is called HDB, his purpose, I do not know his purpose, a purpose because he wanted to launch his own HDB, In addition, because the SQL is not scalable and does not have other cloud characteristics, he wanted to import user-oriented into the SQL, and then failed to try, Amazon was forced to launch RDS.

In other words, you want to use a NoSQL to change the developer's habit of MySQL, as long as your nosql, you need users to change code, there is real cost, then NoSQL will not completely replace the SQL role. So when the SAE launched in 09, it must support SQL, so how to support MySQL? The most important problem we do with MySQL in the cloud is the isolation problem, because the MySQL level is different, we do have some developers on the HCE, even the index does not know what is, built thousands of hundreds of millions of of the table. We do the public cloud computing, if such people are particularly likely to affect our distributed database services, in fact, SQL, or MySQL for the SAE the biggest challenge is isolation. How a person good bad, hacker or his rotten use should not affect other people's use, how to do? is to do it through a virtual machine.

Now the virtual machine technology, should say or relatively mature. For example, I can bind the VCPO to the VPO, of course, network isolation can be done, the actual disk IO isolation Some virtualization can also do, I have a virtual machine up a SQL, user a needs to set up a virtual machine SQL to achieve, this scheme is good. The most important question, the cost of this program is too great, the SAE is very poor, no money, can't afford to. I would like to cite an example, now in the SAE from the current virtualization, a physical machine is up to 30,000 units, more than 30,000 units need 1000 physical machines. I tell you a secret, SAE to now there are no 1000 physical machines, the cost of the SAE is not affordable, we must reduce costs to do segregation.

How to reduce costs? A virtual machine is not a SQL, I have multiple SQL a virtual machine, we are different instance can also be, we have discussed before, in fact, the implementation of this program also has the biggest problem, maintenance of special trouble. You think about that, you have your own lord and the If you manage with managers, you'll go crazy, developers are fine, developers have few things to develop, but the cost of management is very large, SAE how to do, SAE put forward a very crazy concept, so that all users run into a SQL can not, Seems to be a very bad task, but SAE oneself develop a set of products to achieve this technology, is RDC, is the only face public cloud, is to let all users, or a part of users run in a instance, and do not affect each other.

(Responsible editor: The good of the Legacy)