HBase 0.94 版本最近发布了，距上次的0.92 版本发布又四个月了，下面我们就来看一下这个版本中丰富的功能增强。 英文：

中文：http://blog.nosqlfan.com/html/3991.html

性能相关

• 读缓存改进: HDFS 将图片和checksum存储在不同的block中的，所以每次我们进行读操作，都需要进行两次磁盘操作（一次读数据文件，一次读checksum所在的文件）。在 HBASE-5074 这个提案中，提出了将checksum存到block 缓存中的想法。在0.94 版本中，这一特性已经是默认开启的了。这个功能和hfile的新版本有密切的关系。

• 读文件优化: **在HBASE-4465 这个提案中提出了一个叫Lazy Seek的优化，通过先在最近的StoreFile中先读取数据，再看其它StoreFile是否在这个数据操作时间后有操作，如果没有，就不对这个StoreFile进行读取了。这大大减少了对StoreFile的读操作，这一特性也已经是默认开启了。 **Seek optimizations: Till now, if there were several StoreFiles for a column family in a region, HBase would seek in each such files and merge the results, even if the row/column we are looking for is in the most recent file. 在0.94 版本之前，如果一个column family对应了多个StoreFile，那么HBase 会从各个StoreFile中读取数据进行合并，即使我们需要的数据在最近的StoreFile中。 HBASE-4465: “Lazy Seek optimization of StoreFile Scanners” optimizes scanner reads to read the_ most recent_ StoreFile first by lazily seeking the StoreFiles. 在HBASE-4465 这个提案中提出了一个叫Lazy Seek的优化，通过先在最近的StoreFile中先读取数据。This is achieved by introducing a fake keyvalue with its timestamp equal to the maximum timestamp present in the particular StoreFile. 引入一个虚假的keyvalue，该key包含了当前storefile中的最大的时间戳。Thus, a disk seek is avoided until the KeyValueScanner for a StoreFile is bubbled up the heap, implying a need to do a real read operation. This should provide a significant read performance boost, especially for IncrementColumnValue operations where we care only for latest value. This feature is enabled by default.

• WAL文件优化: **The current bottleneck to HBase write speed is replicating the WAL appends across different datanodes. We can speed up this process by compressing the HLog. Current plan involves using a dictionary to compress table name, region id, cf name, and possibly other bits of repeated data. Also, HLog format may be changed in other ways to produce a smaller HLog.详细见HBASE-4608。Adds a custom dictionary-based compression on WAL. **Off by default. To enable, set hbase.regionserver.wal.enablecompression to true in hbase-site.xml.Note that replication is currently broken when WAL compression is enabled.

新特性

• 急求箱工具增强: 对原来的HBck工具进行了增强，提供了对孤儿region、region空洞、重叠region的修复（两个region的row key之间存在重复的key，通过合并成一个新的region）。新的hbck相比以前的改进主要是以hdfs region目录下的.regioninfo为准，而不是.META.。见提案 HBASE-5128: “Uber hbck”。 Online automated repair of table integrity and region consistency problemsHBaseFsck (hbck) has been updated with new repair capabilities. hbck is a tool for checking the region consistency and the table integrity invariants of a running HBase cluster. Checking region consistency verifies that .META., region deployment on region servers and the state of data in HDFS (.regioninfo files) all are in accordance. Table integrity checks verify that all possible row keys resolve to exactly one region of a table – e.g. there are no individual degenerate or backwards regions; no holes between regions; and no overlapping regions. Previously hbck had the ability to diagnose inconsistencies but only had the ability to repair deployment region consistency problems. The updated version now has been augmented with the ability repair region consistency problems in .META. (by patching holes), repair overlapping regions (via merging), patch region holes (by fabricating new regions), and detecting and adopting orphaned regions (by fabricating new .regioninfo file if it is missing in a region’s dir). Caveats:
  • The new hbck selects repairs assuming that HDFS as ground truth, the previous version treated .META. as ground truth.
  • The hbck ‘-fix’ option is present but deprecated and replaced with ‘-fixAssignments’ option.
  • This tool adds APIs in 0.90.7, 0.92.2 and 0.94.0 for clean repairs. The 0.90 version of the tool is compatible with HBase 0.90+, but may require restarting the master or individual individual regionserver for table enable/disable/delete commands to work properly.
• 简化Region Size设置: 通常我们要设置一个region size是很难的，需要考虑很多其它因素。在0.94版本中，按提案 HBASE-4365 的方法，通过动态算法对region size进行制定。
• **更完整的事务支持: **之前Hbase提供行级的事务，不过每次事务只能执行一个写操作，比如连续地执行一系列Put，Delete操作，那么这些操作是单独一个个的事务，其整体并不是原子性执行的。而在0.94版本中，可以实现Put、Delete在同一个事务中一起原子性执行。见提案 HBASE-3584。 实际上上面不过是一些比较大的改进，这一版本共做了140个功能增强，修复了180个bug。这里可以看到更详细的更新。