Hadoop
Discussion of Hadoop. Related subjects include:
Vertica’s version of MapReduce integration
I talked with Omer Trajman of Vertica Monday night about Vertica’s MapReduce integration, part of its Vertica 3.5 release. Highlights included:
- By “integrating Vertica and MapReduce,” Vertica means “integrating Vertica and Hadoop.”
- Vertica’s Hadoop integration is based on Cloudera’s DBInputFormat.
- Omer called out for me several features of Vertica’s Hadoop integration that didn’t just come from Cloudera, namely:
- Cloudera’s DBInputFormat assumes the database runs on a single computer, or a single head node of an MPP system. Vertica’s technology, however, runs on peer parallel nodes with no head, and so Vertica adapted the DBInputFormat technology accordingly.
- Vertica lets you push down Map functions to the database. Omer reports a roughly even division among users and prospects between those who want to do this and ones who don’t.
- Vertica lets you do Reduce functions (or Map functions, if you don’t push them down to the database) on a separate cluster than you run the database software. Vertica asserts that its customers and prospects all want to do this. Right here is the big difference between Vertica’s MapReduce integration and Aster’s or Greenplum’s. (Aster would also say that Vertica’s weaker MapReduce/SQL programming integration is a big difference as well.)
- Indeed, Vertica lets you Reduce into a different DBMS than Vertica, if you choose.
- Vertica gives you flexibility on the size of the Map and Reduce clusters. Omer agreed with me when I said there were some limits on how fast one can add or subtract nodes in a Vertica grid, because there’s data redistribution involved. But one can add/change/delete Hadoop clusters extremely quickly.
Apparently, the use cases for Vertica/Hadoop integration to date lie in algorithmic trading and two kinds of web analytics. Specifically: Read more
NoSQL?
Eric Lai emailed today to ask what I thought about the NoSQL folks, and especially whether I thought their ideas were useful for enterprises in general, as opposed to just Web 2.0 companies. That was the first I heard of NoSQL, which seems to be a community discussing SQL alternatives popular among the cloud/big-web-company set, such as BigTable, Hadoop, Cassandra and so on. My short answers are:
- In most cases, no.
- Most of these technologies are designed for simple, high-volume OLTP (OnLine Transaction Processing.) Most large enterprises have an established way of doing OLTP, probably via relational database management systems. Why change?
- MapReduce is an exception, in that it’s designed for analytics. MapReduce may be useful for enterprises. But where it is, it probably should be integrated into an analytic DBMS.
- There’s one big countervailing factor to all these generalities — schema flexibility.
As for the longer form, let me start by noting that there are two main kinds of reason for not liking SQL. Read more
Facebook’s experiences with compression
One little topic didn’t make it into my long post on Facebook’s Hadoop/Hive-based data warehouse: Compression. The story seems to be:
- Facebook uses gzip, and gets a little bit more than 6X compression.
- Experiments suggest bzip2 would reduce data by another 20% or so, increasing compression to the 7.5X range.
- The downside of bzip2 is 15-25% processing overhead, depending on the kind of data.
| Categories: Data warehousing, Database compression, Facebook, Hadoop | 2 Comments |
Facebook, Hadoop, and Hive
I few weeks ago, I posted about a conversation I had with Jeff Hammerbacher of Cloudera, in which he discussed a Hadoop-based effort at Facebook he previously directed. Subsequently, Ashish Thusoo and Joydeep Sarma of Facebook contacted me to expand upon and in a couple of instances correct what Jeff had said. They also filled me in on Hive, a data-manipulation add-on to Hadoop that they developed and subsequently open-sourced.
Updating the metrics in my Cloudera post,
- Facebook has 400 terabytes of disk managed by Hadoop/Hive, with a slightly better than 6:1 overall compression ratio. So the 2 1/2 petabytes figure for user data is reasonable.
- Facebook’s Hadoop/Hive system ingests 15 terabytes of new data per day now, not 10.
- Hadoop/Hive cycle times aren’t as fast as I thought I heard from Jeff. Ad targeting queries are the most frequent, and they’re run hourly. Dashboards are repopulated daily.
Nothing else in my Cloudera post was called out as being wrong.
In a new-to-me metric, Facebook has 610 Hadoop nodes, running in a single cluster, due to be increased to 1000 soon. Facebook thinks this is the second-largest* Hadoop installation, or else close to it. What’s more, Facebook believes it is unusual in spreading all its apps across a single huge cluster, rather than doing different kinds of work on different, smaller sub-clusters. Read more
| Categories: Data warehousing, EAI, EII, ETL, ELT, ETLT, Facebook, Hadoop, MapReduce, Parallelization, Petabyte-scale data management, Specific users, Web analytics, Yahoo | 48 Comments |
Cloudera presents the MapReduce bull case
Monday was fire-drill day regarding MapReduce vs. MPP relational DBMS. The upshot was that I was quoted in Computerworld and paraphrased in GigaOm as being a little more negative on MapReduce than I really am, in line with my comment
Frankly, my views on MapReduce are more balanced than [my] weary negativity would seem to imply.
Tuesday afternoon the dial turned a couple notches more positive yet, when I talked with Michael Olson and Jeff Hammerbacher of Cloudera. Cloudera is a new company, built around the open source MapReduce implementation Hadoop. So far Cloudera gives away its Hadoop distribution, without charging for any sort of maintenance or subscription, and just gets revenue from professional services. Presumably, Cloudera plans for this business model to change down the road.
Much of our discussion revolved around Facebook, where Jeff directed a huge and diverse Hadoop effort. Apparently, Hadoop played much of the role of an enterprise data warehouse at Facebook — at least for clickstream/network data — including:
- 2 1/2 petabytes of data managed via Hadoop
- 10 terabytes/day of data ingested via Hadoop (Edit: Some of these metrics have been updated in a subsequent post about Facebook.)
- Ad targeting queries run every 15 minutes in Hadoop
- Dashboard roll-up queries run every hour in Hadoop
- Ad-hoc research/analytic Hadoop queries run whenever
- Anti-fraud analysis done in Hadoop
- Text mining (e.g., of things written on people’s “walls”) done in Hadoop
- 100s or 1000s of simultaneous Hadoop queries
- JSON-based social network analysis in Hadoop
Some Facebook data, however, was put into an Oracle RAC cluster for business intelligence. And Jeff does concede that query execution is slower in Hadoop than in a relational DBMS. Hadoop was also used to build the index for Facebook’s custom text search engine.
Jeff’s reasons for liking Hadoop over relational DBMS at Facebook included: Read more
eBay thinks MPP DBMS clobber MapReduce
I talked with Oliver Ratzesberger and his team at eBay last week, who I already knew to be MapReduce non-fans. This time I added more detail.
Oliver believes that, on the whole, MapReduce is 6-8X slower than native functionality in an MPP DBMS, and hence should only be used sporadically. This view is based on part on simulations eBay ran of the Terasort benchmark. On 72 Teradata nodes or 96 lower-powered nodes running another (currently unnamed, as per yet another of my PR fire drills) MPP DBMS, a simulation of Terasort executed in 78 and 120 secs respectively, which is very comparable to the times Google and Yahoo got on 1000 nodes or more.
And by the way, if you use many fewer nodes, you also consume much less floor space or electric power.
| Categories: Analytic technologies, eBay, Hadoop, MapReduce, Parallelization, Teradata | 11 Comments |
Stonebraker, DeWitt, et al. compare MapReduce to DBMS
Along with five other coauthors — the lead author seems to be Andy Pavlo — famous MapReduce non-fans Mike Stonebraker and David DeWitt have posted a SIGMOD 2009 paper called “A Comparison of Approaches to Large-Scale Data Analysis.” The heart of the paper is benchmarks of Hadoop, Vertica, and “DBMS-X” on identical clusters of 100 low-end nodes., across a series of tests including (if I understood correctly):
- A couple of different flavors of a Grep task originally proposed in a Google MapReduce paper.
- A database query on simulated clickstream data
- A join on the same clickstream data.
- Two aggregations on the clickstream data.