Motivation behind Greenplum open source

The biggest news in the past few days is the open source of Greenplum. Pivotal announced the open source of greenplum since the beginning of the year. After more than half a year of waiting, it finally announced the open source at the Postgres Conference in Europe at the end of October, and honored the open source code on Github (https ://github.com/greenplum-db/gpdb ), the official website of the community is http://greenplum.org/, and its official blog also mentions related matters. The code is licensed under the Apache license. Today, some friends in the community have taken down the code from the community and tested the performance of tpch on Greenplum. Although the test is rough, the results are basically the same as the commercial version, which confirms that greenplum will be fully open source. information. Although the key new-generation optimizer orca has not yet seen the code, according to the news, this part will also be opened in the future. It is estimated that this is a cautious operation of this core asset.

The overall impression is that it adopts an open Apache license, and the code is basically open source without reservation. This time, the open source is more a strategic behavior at the company level, rather than a simple tactical behavior of marketing.

Greenplum's open source, in my opinion, has several drivers.

The first is driven by the success of its cloud foundry open source strategy. In terms of the positioning of the entire emc federation, pivotal is a middleware layer, emc is storage, vmware is virtualization, and what pivotal needs to do is PaaS. In terms of pivotal's business, the two pillars of cloud and big data must be completed in order to stabilize the territory of the emc federation. In the cloud layer, vmware is already the overlord of the private cloud, and if the remaining upper layer PaaS layer falls into the hands of others, it is also a big threat to it, so the PaaS layer is also determined to win. Given that there were already many competitors in the market at that time, cloudfoundry started in 2011 with an open source model, pulling IBM and HP up. This strategy has achieved unexpectedly great success, and now CF has almost become the de facto standard for PaaS, IBM has also launched a CF-based bluemix product, and pivotal released its 2014 financial report this year, with a very eye-catching title: record breaking In 2014, the fastest ever open source product sales growth ( http://finance.yahoo.com/news/pivotal-cloud-foundry-reports-record-160000128.html ) gained 4,000 in just one year 10,000 US dollars in software sales revenue, and 100 customers in the Fortune 500! This is basically a myth for basic software. Having said so much, it is nothing more than to emphasize to everyone that, under the great success of cloudfoudry, the entire pivotal understanding of open source has reached the level of business strategy, there is no need to discuss whether to open source or not, open source has become a killer weapon ! In this context, it is easier for us to understand the open source of greenplum.

Second, Pivotal's big data battlefield requires new strategic adjustments. As mentioned above, in addition to cloud, Pivotal's strategic focus is big data. In addition to its own hadoop distribution, Pivotal also has greenplum as the most important asset of big data. However, the whole big data market is not very ideal. In addition to the three third-party distributions of Cloudera, Hortonworks, and MapR, there are also distributions of Pivotal, IBM, and Intel among Hadoop distributions. The most prominent problem in the market is fragmentation. Fragmented, the threshold for Hadoop is low, there are many manufacturers, and the homogeneity is serious. Not only is it difficult to sell at a premium, but the way for traditional big manufacturers to sell basic software at a premium to obtain excess profits begins to fail; more importantly, more and more of customers tend to use third-party independent distributions. The reason is not difficult to understand. Users prefer that there are distribution vendors such as Redhat and SuSe in the Linux ecosystem, rather than vendor-controlled ecosystems like AIX or Windows Server. , which further exacerbates the predicament faced by manufacturers. Due to the poor market performance of its distribution, Intel began to disband its own Hadoop team one after another. The domestic Star Ring was born under this background. In 2014, Intel invested heavily in Cloudera 7. 400 million US dollars to obtain 18% of the equity, Dell also invested in the Cloudera camp, HP, Microsoft, Teradata tend to directly use the Hortonworks distribution, HP also invested in Hortonworks. However, Cloudera has obvious advantages. Its revenue is nearly double that of Hortonworks. Cloudera seems to be on the verge of becoming the new overlord of the Hadoop ecosystem. The market structure is slowly solidifying, and Intel may be the winner of this battle. In such a market structure, the predicament of Pivotal can be imagined. The HD distribution must need to readjust its strategy. Therefore, Pivotal, together with 15 disadvantaged players such as IBM and GE, announced the establishment of the ODP (Open Data Platform) organization in March this year. , essentially hoping to strike a balance against Cloudera by supporting Hortonworks. But the odds of this bet are not obvious, and Cloudera's CEO even publicly mocked ODP, saying that its appearance itself is a victory for Cloudera (https://gigaom.com/2015/03/03/cloudera-ceo-declares-victory-over -big-data-competition/ ), the open source of Greenplum appeared as the weight of ODP, Pivotal decided to open source HD distribution, Gemfire, HAWQ and Greenplum, so from this perspective, we can see that the open source of Greenplum itself, It is a weight that Pivotal hopes to win back this battle. For it, the more people use it, the better, and there is no need to hide any functions.

In fact, before Greenplum was open-sourced, the MPP database was not easy, and the market was fragmented. Each manufacturer only had a revenue of tens of millions of dollars. It was difficult to make a big breakthrough in the market, and it was difficult to enter the traditionally rich DW market. , The Hadoop ecosystem is mainly open source, and even many products like Impala are doing their similar functions. The traditional ones cannot be opened, and it is difficult to fully open new markets, and there is a lot of competition and a dilemma. And Greenplum's architecture for more than a decade is unable to make major adjustments. It is in this context that it is better to open source and revitalize the overall situation. The following article calls the move open sourcing code is the modern graceful way to retire an unprofitable product line (http://skylandtech.net/2015/02/24/thinking-about-the-pivotal-announcements/ )

by Greenplum Open source should be a relatively aggressive and aggressive move. Success or failure is unpredictable, but it will have a relatively large impact on the entire ecosystem.