小公司颠覆存储业大野心：冲击百亿美元数据库大市场

    可以肯定的是，如今不管哪款企业级应用，一定都会用到关系型数据库管理系统（RDBMS）。自从20世纪80年代以来，无论信息类型是关于金融、小插件，还是人员，RDBMS就一直是存储及恢复几乎所有类型信息的首选方案。甲骨文公司（Oracle）、IBM公司和微软公司（Microsoft）无疑是业界翘楚，SAP公司和天睿公司（Teradata）也是强劲的竞争者。

    随后大数据横空出世。爆炸式增长的海量新数据使目前RDBMS的局限一目了然，也着实恼人。风投公司InterWest Partners普通合伙人布鲁斯•克利夫兰称：“以前企业可能花上几分钟、几小时、甚至好几天来回答客户的问题，但今天，客户希望立即马上就有答复。”

    在面向消费者的行业里，客户可能想了解“有现货吗？”而在B2B行业里，客户可能想知道“最优惠的价格是多少？”而且他们希望马上就能获得答案。对此克利夫兰补充称：“随着数据量的猛增，企业现在只有两个次优选择：要么维持欠佳的业绩，要么用极其昂贵的硬件‘升级’传统数据库。”

    Splice Machine正是抓住了这一契机。就像IBM公司最近宣布要“向外扩展”存储产品，让大公司无需巨额投资也能升级存储能力一样，Splice Machine也瞄准了大型数据库提供类似服务。

两全其美的方案

    作为目前RDBMS的替代方案，Splice Machine有效地将传统关系型数据技术与Hadoop向外扩展的能力相结合，而后者实际上正是雅虎公司（Yahoo）、Facebook公司及其他大企业大数据架构的标准。尽管Hadoop以执行批量分析而著称，但Splice Machine主攻的却是业务应用和实时分析。

    咨询公司Pund-IT首席分析师查尔斯•金称：“Splice Machine的目标是颠覆传统观念，提供一种充分结合两大领域优势的解决方案，也就是：Hadoop的经济实惠、向外扩展的性能，及其与通用RDBMS/SQL操作——如数据清理、客户上卷和提取转换加载（ETL）——以及像Cognos, Unica, SAS和SPSS这类应用的良好兼容性。”

    克利夫兰表示，Splice Machine的目标是，凭借高效能、精确性和可靠性等承诺，说服企业客户用他们的产品替换掉MySQL和Oracle这类传统数据库。

    他说：“（企业）可以通过将数据向外扩展到相对廉价的服务器集群中获得高效能。同时，由于Splice Machine支持原生结构化查询语言（SQL），企业可以无须重写应用就能实现这一点。”

“前所未有的可扩展性”

    Splice Machine创立于两年前，获得了来自InterWest和Mohr Davidow Ventures两家公司共计1900万美元的投资。上周，Splice Machine发布了基于Hadoop的公测版数据库。已有15家公司采用了这项技术，包括德克萨斯州圣安东尼奥的营销服务公司Harte Hanks。Splice Machine称，Harte Hanks发现与其现有的Oracle RAC数据库相比，系统的性价比提高了十倍多（Splice Machine提供的用于测试的标准版数据库免费；否则每个节点收费5,000美元）。

    Splice Machine联合创始人兼首席执行官蒙特•兹魏本告诉《财富》杂志（Fortune）：“我们认为这款产品将颠覆价值210亿美元的数据库市场。”【高德纳公司（Gartner）则估计该市场价值为260亿美元，年增长率为9%】。

    Name any enterprise application today, and it's a pretty safe bet that it's deployed on a relational database management system. Since the 1980s, RDBMS technology has been the go-to solution for storing and retrieving information about almost anything, from finances to widgets to people. Oracle (ORCL), IBM (IBM), and Microsoft (MSFT) are among the companies leading the way; SAP and Teradata (TDC) are also competitors.

    Then big data came along. Newly exploding volumes of data made the limits of existing RDMBSes painfully clear. "Buyers now want, and expect, immediate answers to questions that previously may have taken companies minutes, hours, days to answer," said Bruce Cleveland, a general partner with InterWest Partners.

    In a consumer-facing business, a buyer may want to know, "Is it in stock?" In a B2B business, the buyer may want to know, "What is your best price?" And they want to know immediately. "With data volumes increasing, companies have had two poor options: Either live with sub-optimal performance or 'scale up' their traditional databases using extremely expensive hardware," Cleveland added.

    That's where Splice Machine comes in. Much the way IBM recently announced a "scale out" storage offering that allows large companies to scale capabilities without massive investments, so Splice Machine aims to do the same thing for large databases.

'The best of both worlds'

    An alternative to today's RDBMSes, Splice Machine effectively combines traditional relational database technology with the scale-out capabilities of Hadoop, the de facto standard for big data architecture at Yahoo (YHOO), Facebook (FB), and most other major companies. Whereas Hadoop is best known for its power to perform batch analytics, Splice Machine focuses instead on operational applications and real-time analytics.

    "Splice Machine is aiming to turn the common wisdom on its head with a solution that combines the best of both worlds: The cost-effective, scale-out performance of Hadoop and compatibility with common RDBMS/SQL processes -- such as data cleansing, customer roll-ups and ETL work -- and applications like Cognos, Unica, SAS and SPSS," Charles King, a principal analyst withPund-IT, said.

    Splice Machine's aim is to convince businesses to replace traditional databases such as MySQL and Oracle with their own, Cleveland said, with promises of high performance, accuracy, and reliability.

    "[Businesses] get high performance from scaling out onto clusters of inexpensive servers," he said. "And because Splice Machine supports native SQL, companies can do this without having to rewrite their applications."

'Scalability nobody has been addressing'

    Last week, Splice Machine -- which was founded two years ago and is backed by $19 million in funding from InterWest Partners and Mohr Davidow Ventures -- launched its Hadoop-based database into public beta. Fifteen companies already use the technology, including the San Antonio, Texas-based marketing services company Harte Hanks. Splice Machine said Harte Hanks observed improvement to the ratio of price to performance by more than 10x compared with its existing Oracle RAC databases. (Splice Machine offers the standalone version of its database on a freemium basis for testing; otherwise, it's $5,000 per node.)

    "We think this will disrupt the $21 billion database market," Monte Zweben, Splice Machine's co-founder and chief executive, told Fortune. (Gartner estimates the market to be worth $26 billion and growing at nine percent annually.)