Database compression
Analysis of technology that compresses data within a database management system. Related subjects include:
Big stuff coming from DATAllegro
In the literal sense, that is. While the details on what I wrote about this a few weeks ago* are still embargoed, I’m at liberty to drop a few more hints.
*Please also see DATAllegro CEO Stuart Frost’s two comments added today to that thread.
DATAllegro systems these days basically consist of Dell servers talking to EMC disk arrays, with Cisco Infiniband to provide fast inter-server communication without significant CPU load. Well, if you decrease the number of Dell servers per EMC box, and increase the number of disks per EMC box, you can slash your per-terabyte price (possibly at the cost of lowering performance).
Read more
| Categories: Data warehouse appliances, Data warehousing, Database compression, DATAllegro | Leave a Comment |
DATAllegro heads for the high end
DATAllegro Stuart Frost called in for a prebriefing/feedback/consulting session. (I love advising my DBMS vendor clients on how to beat each other’s brains in. This was even more fun in the 1990s, when combat was generally more aggressive. Those were also the days when somebody would change jobs to an arch-rival and immediately explain how everything they’d told me before was utterly false …)
While I had Stuart on the phone, I did manage to extract some stuff I’m at liberty to use immediately. Here are the highlights: Read more
| Categories: Data warehouse appliances, Data warehousing, Database compression, DATAllegro, Greenplum, Netezza, Teradata | 4 Comments |
Fast RDF in specialty relational databases
When Mike Stonebraker and I discussed RDF yesterday, he quickly turned to suggesting fast ways of implementing it over an RDBMS. Then, quite characteristically, he sent over a paper that allegedly covered them, but actually was about closely related schemes instead. 🙂 Edit: The paper has a new, stable URL. Hat tip to Daniel Abadi.
All minor confusion aside, here’s the story. At its core, an RDF database is one huge three-column table storing subject-property-object triples. In the naive implementation, you then have to join this table to itself repeatedly. Materialized views are a good start, but they only take you so far. Read more
| Categories: Columnar database management, Data models and architecture, Data warehousing, Database compression, RDF and graphs, Theory and architecture, Vertica Systems | 1 Comment |
The petabyte machine
EMC has announced a machine — a virtual tape library — that supposedly stores 1.8 petabytes of data. Even though that’s only 584 terabytes uncompressed, it shows that the 1 petabyte barrier will be broken soon no matter how unhyped the measurement.
I just recently encountered some old notes in which Sybase proudly announced a “1 gigabyte challenge.” The idea was that 1 gig was a breakthrough size for business databases.
| Categories: Database compression, EMC, Sybase, Theory and architecture | Leave a Comment |
Will database compression change the hardware game?
I’ve recently made a lot of posts about database compression. 3X or more compression is rapidly becoming standard; 5X+ is coming soon as processor power increases; 10X or more is not unrealistic. True, this applies mainly to data warehouses, but that’s where the big database growth is happening. And new kinds of data — geospatial, telemetry, document, video, whatever — are highly compressible as well.
This trend suggests a few interesting possibilities for hardware, semiconductors, and storage.
- The growth in demand for storage might actually slow. That said, I frankly think it’s more likely that Parkinson’s Law of Data will continue to hold: Data expands to fill the space available. E.g., video and other media have near-infinite potential to consume storage; it’s just a question of resolution and fidelity.
- Solid-state (aka semiconductor or flash) persistent storage might become practical sooner than we think. If you really can fit a terabyte of data onto 100 gigs of flash, that’s a pretty affordable alternative. And by the way — if that happens, a lot of what I’ve been saying about random vs. sequential reads might be irrelevant.
- Similarly, memory-centric data management is more affordable when compression is aggressive. That’s a key point of schemes such as SAP’s or QlikTech’s. Who needs flash? Just put it in RAM, persisting it to disk just for backup.
- There’s a use for faster processors. Compression isn’t free. What you save on disk space and I/O you pay for at the CPU level. Those 5X+ compression levels do depend on faster processors, at least for the row store vendors.
| Categories: Data warehousing, Database compression, Memory-centric data management, QlikTech and QlikView, SAP AG | 6 Comments |
Mike Stonebraker on database compression — comments
In my opinion, the key part of Mike Stonebraker’s fascinating note on data compression was (emphasis mine):
The standard wisdom in most row stores is to use block compression. Hence, a storage block is compressed using a single technique (say Lempel-Ziv or dictionary). The technique chosen then compresses all the attributes in all the columns which occur on the block. In contrast, Vertica compresses a storage block that only contains one attribute. Hence, it can use a different compression scheme for each attribute. Obviously a compression scheme that is type-specific will beat an implementation that is “one size fits all”.
It is possible for a row store to use a type-specific compression scheme. However, if there are 50 attributes in a record, then it must remember the state for 50 type-specific implementations, and complexity increases significantly.
In addition, all row stores we are familiar with decompress each storage block on access, so that the query executor processes uncompressed tuples. In contrast, the Vertica executor processes compressed tuples. This results in better L2 cache locality, less main memory copying and generally much better performance.
Of course, any row store implementation can rewrite their executor to run on compressed data. However, this is a rewrite – and a lot of work.
| Categories: Columnar database management, Data warehousing, Database compression, Sybase, Vertica Systems | 8 Comments |
Mike Stonebraker explains column-store data compression
The following is by Mike Stonebraker, CTO of Vertica Systems, copyright 2007, as part of our ongoing discussion of data compression. My comments are in a separate post.
Row Store Compression versus Column Store Compression
I Introduction
There are three aspects of space requirements, which we discuss in this short note, namely:
structural space requirements
index space requirements
attribute space requirements.
| Categories: Data warehousing, Database compression, Michael Stonebraker, Theory and architecture, Vertica Systems | 7 Comments |
Compression in columnar data stores
We have lively discussions going on columnar data stores vs. vertically partitioned row stores. Part is visible in the comment thread to a recent post. Other parts come in private comments from Stuart Frost of DATAllegro and Mike Stonebraker of Vertica et al.
To me, the most interesting part of what the Vertica guys are saying is twofold. One is that data compression just works better in column stores than row stores, perhaps by a factor of 3, because “the next thing in storage is the same data type, rather than a different one.” Frankly, although Mike has said this a couple of times, I haven’t understood yet why row stores can’t be smart enough to compress just as well. Yes, it’s a little harder than it would be in a columnar system; but I don’t see why the challenge would be insuperable.
The second part is even cooler, namely the claim that column stores allow the processors to operate directly on compressed data. But once again, I don’t see why row stores can’t do that too. For example, when you join via bitmapped indices, exactly what you’re doing is operating on highly-compressed data.
| Categories: Columnar database management, Data warehouse appliances, Data warehousing, Database compression, DATAllegro, Vertica Systems | 2 Comments |
Word of the day: “Compression”
IBM sent over a bunch of success stories recently, with DB2’s new aggressive compression prominently mentioned. Mike Stonebraker made a big point of Vertica’s compression when last we talked; other column-oriented data warehouse/mart software vendors (e.g. Kognitio, SAP, Sybase) get strong compression benefits as well. Other data warehouse/mart specialists are doing a lot with compression too, although some of that is governed by please-don’t-say-anything-good-about-us NDA agreements.
Compression is important for at least three reasons:
- It saves disk space, which is a major cost issue in data warehousing.
- It saves I/O, which is the major performance issue in data warehousing.
- In well-designed systems, it can actually make on-chip execution faster, because the gains in memory speed and movement can exceed the cost of actually packing/unpacking the data. (Or so I’m told; I haven’t aggressively investigated that claim.)
When evaluating data warehouse/mart software, take a look at the vendor’s compression story. It’s important stuff.
EDIT: DATAllegro claims in a note to me that they get 3-4x storage savings via compression. They also make the observation that fewer disks ==> fewer disk failures, and spin that — as it were 🙂 — into a claim of greater reliability.
| Categories: Data warehouse appliances, Data warehousing, Database compression, DATAllegro, IBM and DB2, SAP AG, Vertica Systems | 3 Comments |
Are row-oriented RDBMS obsolete?
If Mike Stonebraker is to be believed, the era of columnar data stores is upon us.
Whether or not you buy completely into Mike’s claims, there certainly are cool ideas in his latest columnar offering, from startup Vertica Systems. The Vertica corporate site offers little detail, but Mike tells me that the product’s architecture closely resembles that of C-Store, which is described in this November, 2005 paper.
The core ideas behind Vertica’s product are as follows. Read more