We all tend to assume that data is a great and glorious asset. How solid is this assumption?
- Yes, data is one of the most proprietary assets an enterprise can have. Any of the Goldman Sachs big three* — people, capital, and reputation — are easier to lose or imitate than data.
- In many cases, however, data’s value diminishes quickly.
- Determining the value derived from owning, analyzing and using data is often tricky — but not always. Examples where data’s value is pretty clear start with:
- Industries which long have had large data-gathering research budgets, in areas such as clinical trials or seismology.
- Industries that can calculate the return on mass marketing programs, such as internet advertising or its snail-mail predecessors.
*”Our assets are our people, capital and reputation. If any of these is ever diminished, the last is the most difficult to restore.” I love that motto, even if Goldman Sachs itself eventually stopped living up to it. If nothing else, my own business depends primarily on my reputation and information.
This all raises the idea – if you think data is so valuable, maybe you should get more of it. Areas in which enterprises have made significant and/or successful investments in data acquisition include:
- Actual scientific, clinical, seismic, or engineering research.
- Actual selling of (usually proprietary) data, with the straightforward economic proposition of “Get once, sell to multiple customers more cheaply than they could get it themselves.” Examples start:
- This is the essence of the stock quote business. And Michael Bloomberg started building his vast fortune by adding additional data to what the then-incumbents could offer, for example by getting fixed-income prices from Cantor Fitzgerald.*
- Multiple marketing-data businesses operate on this model.
- Back when there was a small but healthy independent paper newsletter and directory business, its essence was data.
- And now there are many online data selling efforts, in niches large and small.
- Internet ad-targeting businesses. Making money from your great ad-targeting technology usually involves access to lots of user-impression and de-anonymization data as well.
- Aggressive testing by internet businesses, of substantive offers and marketing-display choices alike. At the largest, such as eBay, you’ll rarely see a page that doesn’t have at least one experiment on it. Paper-based direct marketers take a similar approach. Call centers perhaps should follow suit more than they do.
- Surveys, focus groups, etc. These are commonly expensive and unreliable (and the cheap internet ones commonly irritate people who do business with you). But sometimes they are, or seem to be, the only kind of information available.
- Free-text data. On the whole I’ve been disappointed by the progress in text analytics. Still — and this overlaps with some previous points — there’s a lot of information in text or narrative form out there for the taking.
- Internally you might have customer emails, call center notes, warranty reports and a lot more.
- Externally there’s a lot of social media to mine.
*Sadly, Cantor Fitzgerald later became famous for being hit especially hard on 9/11/2001.
And then there’s my favorite example of all. Several decades ago, especially in the 1990s, supermarkets and mass merchants implemented point-of-sale (POS) systems to track every item sold, and then added loyalty cards through which they bribed their customers to associate their names with their purchases. Casinos followed suit. Airlines of course had loyalty/frequent-flyer programs too, which were heavily related to their marketing, although in that case I think loyalty/rewards were truly the core element, with targeted marketing just being an important secondary benefit. Overall, that’s an awesome example of aggressive data gathering. But here’s the thing, and it’s an example of why I’m confused about the value of data — I wouldn’t exactly say that grocers, mass merchants or airlines have been bastions of economic success. Good data will rarely save a bad business.
Everybody is confused about privacy and surveillance. So I’m renewing my efforts to consciousness-raise within the tech community. For if we don’t figure out and explain the issues clearly enough, there isn’t a snowball’s chance in Hades our lawmakers will get it right without us.
How bad is the confusion? Well, even Edward Snowden is getting it wrong. A Wired interview with Snowden says:
“If somebody’s really watching me, they’ve got a team of guys whose job is just to hack me,” he says. “I don’t think they’ve geolocated me, but they almost certainly monitor who I’m talking to online. Even if they don’t know what you’re saying, because it’s encrypted, they can still get a lot from who you’re talking to and when you’re talking to them.”
That is surely correct. But the same article also says:
“We have the means and we have the technology to end mass surveillance without any legislative action at all, without any policy changes.” The answer, he says, is robust encryption. “By basically adopting changes like making encryption a universal standard—where all communications are encrypted by default—we can end mass surveillance not just in the United States but around the world.”
That is false, for a myriad of reasons, and indeed is contradicted by the first excerpt I cited.
What privacy/surveillance commentators evidently keep forgetting is:
- There are many kinds of privacy-destroying information. I think people frequently overlook just how many kinds there are.
- Many kinds of organization capture that information, can share it with each other, and gain benefits from eroding or destroying privacy. Similarly, I think people overlook just how pervasive the incentive is to snoop.
- Privacy is invaded through a variety of analytic techniques applied to that information.
So closing down a few vectors of privacy attack doesn’t solve the underlying problem at all.
Worst of all, commentators forget that the correct metric for danger is not just harmful information use, but chilling effects on the exercise of ordinary liberties. But in the interest of space, I won’t reiterate that argument in this post.
Perhaps I can refresh your memory why each of those bulleted claims is correct. Major categories of privacy-destroying information (raw or derived) include:
- The actual content of your communications – phone calls, email, social media posts and more.
- The metadata of your communications — who you communicate with, when, how long, etc.
- What you read, watch, surf to or otherwise pay attention to.
- Your purchases, sales and other transactions.
- Video images, via stationary cameras, license plate readers in police cars, drones or just ordinary consumer photography.
- Monitoring via the devices you carry, such as phones or medical monitors.
- Your health and physical state, via those devices, but also inferred from, for example, your transactions or search engine entries.
- Your state of mind, which can be inferred to various extents from almost any of the other information areas.
- Your location and movements, ditto. Insurance companies also want to put monitors in cars to track your driving behavior in detail.
Of course, these categories overlap. For example, information about your movements can be derived not just from your mobile phone, but also from your transactions, from surveillance cameras, and from the health-monitoring devices that are likely to become much more pervasive in the future.
So who has reason to invade your privacy? Unfortunately, the answer boils down to “just about everybody”. In particular:
- Any internet or telecom business would like to know, in great detail, what you are doing with their offerings, along with any other information that might influence what you’re apt to buy or do next.
- Anybody who markets or sells to consumers wants to know similar things.
- Similar things are true of anybody who worries about credit or insurance risk.
- Anybody who worries about fraud wants to know who you’re connected to, and also wants to match you against any known patterns of fraud-related behavior.
- Anybody who hires employees wants to know who might be likely to work hard, get sick or quit.
- Similarly, they’d like to know who does or might engage in employee misconduct.
- Medical researchers and caregivers have some of the most admirable reasons for wanting to violate privacy.
And that’s even without mentioning the most obvious suspects — law enforcement and national security of many kinds, who can be presumed to in at least certain cases be able to get any information that’s available to any other organization.
Finally, my sense is:
- People appreciate the potential of fancy-schmantzy language and image recognition.
- The graph analysis done on telecom metadata is so simple that people generally “get” what’s going on.
- Despite all the “big data analytics” hype, commentators tend to forget just how powerful machine learning/predictive analytics privacy intrusions could be. Those psychographic clustering techniques devised to support advertising and personalization could be applied in much more sinister ways as well.
- The crucial point about chilling effects was laid out in two July, 2013 posts, and some public policy recommendations around the same time. Those four posts are a great starting point for the non-technical “bottom line” part of the discussion. A January, 2014 post adds some more political context.
- A January, 2011 post on the technology of privacy threats adds detail to many of the points above.
- A February, 2014 post on various metadata-related confusions notes some egregious governmental spin.
I’ve talked with many companies recently that believe they are:
- Focused on building a great data management and analytic stack for log management …
- … unlike all the other companies that might be saying the same thing …
- … and certainly unlike expensive, poorly-scalable Splunk …
- … and also unlike less-focused vendors of analytic RDBMS (which are also expensive) and/or Hadoop distributions.
At best, I think such competitive claims are overwrought. Still, it’s a genuinely important subject and opportunity, so let’s consider what a great log management and analysis system might look like.
Much of this discussion could apply to machine-generated data in general. But right now I think more players are doing product management with an explicit conception either of log management or event-series analytics, so for this post I’ll share that focus too.
A short answer might be “Splunk, but with more analytic functionality and more scalable performance, at lower cost, plus numerous coupons for free pizza.” A more constructive and bottoms-up approach might start with:
- Agents for any kind of machine that admits streams of data.
- Parsers that:
- Immediately identify explicit name-value pairs in popular formats such as JSON or XML.
- Also immediately extract a significant fraction of all implicit fields in text strings — timestamps for sure, but also a lot else. (Splunk is the current gold standard for such capabilities.)
- Allow you to easily write rules for more such extractions.
- Immediate indexing in line with everything the parsers do.
- Easy import of log files, relational tables, and other relevant data structures.
- Queries that can exploit all the indexes, at least up to the functionality level of SQL 2003 analytics (including windowing) and StreamSQL, of course with …
- … blazing scalable performance.
- Strong workload management and concurrent performance support. (Teradata is the gold standard for such capabilities in the analytic sphere.)
- Various other mature-DBMS features, e.g. in backup, manageability, and uptime.
Further, there would be numerous styles of business intelligence interface, at least including:
- Generic BI like we generally see for tabular data.
- Constantly-changing displays of streaming data.
- BI with an event-series orientation.
- Strong alerting.
- Mobile versions of everything.
The data management part of that is particularly hard, in that:
- Different architectures seem naturally well-suited for different parts of the problem.
- Maturing a new data management product is always difficult, costly and slow.
My thoughts on strengths and weaknesses of some obvious log data management contenders start:
- Oracle, IBM, and Microsoft have a lot of heft in all things database. But while each of those vendors has great resources and occasionally impressive pieces of new database engineering, none shows much evidence of framing, let alone solving, the problem in the right way(s).
- SAP owns Sybase, HANA, several old CEP companies, and Business Objects. Add them to the Oracle/IBM/Microsoft list.
- Teradata has a lot going for them. Their core analytic data management strengths are obvious. They’ve owned Aster for a while, and Aster innovated nPath quite some time ago. They recently added Hadapt, a leader in schema-on-need, as well as Revelytix, which has some good ideas in dataset management. Like most other DBMS vendors, however, Teradata doesn’t yet have much of a story for streaming data, and anyhow the most optimistic case for Teradata involves the difficult task of stitching together disparate data management technologies.
- HP Vertica has a decent position as well. Probably more proven in general concurrent, scalable performance than others in their peer group (Netezza, Greenplum, et al.), Vertica also was relatively early in innovations relevant to log analysis, including a range of time series/event series features and its own schema-on-need effort. Vertica was also founded by people who were also streaming pioneers (there were heavily overlapping groups of academics behind StreamBase, Vertica and VoltDB), but it’s not clear how that background is reflected in present Vertica product.
- Splunk, of course, has a complete stack. At the data acquisition and parsing layers, it’s second to none, and it has a considerable set of log-appropriate BI capabilities as well. And for data management it in effect is stitching together two different inverted-list data stores, plus Hadoop.
- Hadoop distribution vendors such as Cloudera, MapR or Hortonworks offer typically bundle a range of relevant capabilities. HDFS (Hadoop Distributed File System) is the default place to dump entire logs. In most distros, Spark offers a new approach to streaming. Impala, Drill and so on offer query. Flume gathers the log data in the first place. But a lot of the cooler capabilities are immature or unproven, and in some cases that’s putting it mildly.
In the interest of length, I’ll omit discussion of smaller vendors, except to say that Platfora’s integrated-stack event series analytics story deserves attention, and I’m disappointed that I never hear about Sumo Logic. And I don’t know a lot about companies positioned as SIEM (Security Information and Event Management), especially now that SenSage has left the scene.
I spent a day with Teradata in Rancho Bernardo last week. Most of what we discussed is confidential, but I think the non-confidential parts and my general impressions add up to enough for a post.
First, let’s catch up with some personnel gossip. So far as I can tell:
- Scott Gnau runs most of Teradata’s development, product management, and product marketing, the big exception being that …
- … Darryl McDonald run the apps part (Aprimo and so on), and no longer is head of marketing.
- Oliver Ratzesberger runs Teradata’s software development.
- Jeff Carter has returned to his roots and runs the hardware part, in place of Carson Schmidt.
- Aster founders Mayank Bawa and Tasso Argyros have left Teradata (perhaps some earn-out period ended).
- Carson is temporarily running Aster development (in place of Mayank), and has some sort of evangelism role waiting after that.
- With the acquisition of Hadapt, Teradata gets some attention from Dan Abadi. Also, they’re retaining Justin Borgman.
The biggest change in my general impressions about Teradata is that they’re having smart thoughts about the cloud. At least, Oliver is. All details are confidential, and I wouldn’t necessarily expect them to become clear even in October (which once again is the month for Teradata’s user conference). My main concern about all that is whether Teradata’s engineering team can successfully execute on Oliver’s directives. I’m optimistic, but I don’t have a lot of detail to support my good feelings.
In some quick-and-dirty positioning and sales qualification notes, which crystallize what we already knew before:
- The Teradata 1xxx series is focused on cost-per-bit.
- The Teradata 2xxx series is focused on cost-per-query. It is commonly Teradata’s “lead” product, at least for new customers.
- The Teradata 6xxx series is supposed to be above to do “everything”.
- The Teradata Aster “Discovery Analytics” platform is sold mainly to customers who have a specific high-value problem to solve. (Randy Lea gave me a nice round dollar number, but I won’t share it.) I like that approach, as it obviates much of the concern about “Wait — is this strategic for us long-term, given that we also have both Teradata database and Hadoop clusters?”
- 1xxx and 2xxx systems are meant to be I/O-constrained. 6xxx systems are meant to be constrained mainly by CPU, but every system will be I/O-constrained at some point.
- There is at least one example of a Very Well Known organization buying Teradata’s Hadoop-only appliance despite not otherwise being a Hadoop customer. Teradata concedes, however, that this is not a common occurrence.
- Customers are increasingly using co-location rather than their own data centers. Many colo organizations charge more or less strictly by floor space. Hence, there’s a push for maximum processing density per rack, power density and weight be damned.
Speaking of not being CPU-constrained — I heard 7-10% as an estimate for typical Hadoop utilization, and also 10-15%. While I didn’t ask, I presume these figures assume traditional MapReduce types of Hadoop workloads. I’m not sure why these figures are yet lower than eBay’s long-ago estimates of Hadoop “parallel efficiency”.
Like Carson used to do, Jeff shared a variety of hardware and networking tidbits with me. In particular:
- Jeff is confident in Moore’s Law continuing for at least 5 more years. (I think that’s a near-consensus; the 2020s, however, are another matter.)
- Teradata still uses SAS rather than SATA for all disk (spinning or solid-state) controllers. They’re now seeing 6-700 MB/sec/device on SSDs (Solid State Disk), up from 3-400.
- SSD prices are down 60% over the past 6 months, vs. much slower declines previously.
- Formerly a SanDisk/Pliant partisan, Teradata now thinks there are multiple vendors of good SSDs. (I’m not sure whether they’d be happy if I said which one they currently like best.)
- Jeff foresees InfiniBand and Ethernet more or less merging. Right now Teradata is using a lot of 56 Gb/sec InfiniBand.
Since Oliver is now a Teradata mucky-muck, I asked about virtual data marts, an idea that he pretty much invented or at least popularized back in his eBay days. Comments included:
- Teradata now calls them Data Labs.
- Adoption is very high.
- One major feature is “time boxing” — they expire after a period of time unless you renew them.
- Analysis of virtual data mart usage is a good guide as to what you might want to add to your permanent data warehouse.
And I’ll stop here, although I hope that a couple more-focused posts will also eventually flow from the visit.
Many of the companies I talk with boast of freeing business analysts from reliance on IT. This, to put it mildly, is not a unique value proposition. As I wrote in 2012, when I went on a history of analytics posting kick,
- Most interesting analytic software has been adopted first and foremost at the departmental level.
- People seem to be forgetting that fact.
In particular, I would argue that the following analytic technologies started and prospered largely through departmental adoption:
- Fourth-generation languages (the analytically-focused ones, which in fact started out being consumed on a remote/time-sharing basis)
- Electronic spreadsheets
- 1990s-era business intelligence
- Fancy-visualization business intelligence
- Predictive analytics
- Text analytics
- Rules engines
What brings me back to the topic is conversations I had this week with Paxata and Metanautix. The Paxata story starts:
- Paxata is offering easy — and hopefully in the future comprehensive — “data preparation” tools …
- … that are meant to be used by business analysts rather than ETL (Extract/Transform/Load) specialists or other IT professionals …
- … where what Paxata means by “data preparation” is not specifically what a statistician would mean by the term, but rather generally refers to getting data ready for business intelligence or other analytics.
Metanautix seems to aspire to a more complete full-analytic-stack-without-IT kind of story, but clearly sees the data preparation part as a big part of its value.
If there’s anything new about such stories, it has to be on the transformation side; BI tools have been helping with data extraction since — well, since the dawn of BI. The data movement tool I used personally in the 1990s was Q+E, an early BI tool that also had some update capabilities.* And this use of BI has never stopped; for example, in 2011, Stephen Groschupf gave me the impression that a significant fraction of Datameer’s usage was for lightweight ETL.
*Q+E came from Pioneer Software, the original predecessor of Progress DataDirect, which first came to fame in association with Microsoft Excel and the invention of ODBC.
More generally, I’d say that there are several good ways for IT to give out data access, the two most obvious of which are:
- “Semantic layers” in BI tools.
- Data copies in departmental data marts.
If neither of those works for you, then most likely either:
- Your problem isn’t technology.
- Your problem isn’t data access.
And so we’ve circled back to what I wrote last month:
Data transformation is a better business to enter than data movement. Differentiated value in data movement comes in areas such as performance, reliability and maturity, where established players have major advantages. But differentiated value in data transformation can come from “intelligence”, which is easier to excel in as a start-up.
What remains to be seen is whether and to what extent any of these startups (the ones I mentioned above, or Trifacta, or Tamr, or whoever) can overcome what I wrote in the same post:
When I talk with data integration startups, I ask questions such as “What fraction of Informatica’s revenue are you shooting for?” and, as a follow-up, “Why would that be grounds for excitement?”
It will be interesting to see what happens.
I have a small blacklist of companies I won’t talk with because of their particularly unethical past behavior. Actian is one such; they evidently made stuff up about me that Josh Berkus gullibly posted for them, and I don’t want to have conversations that could be dishonestly used against me.
That said, Peter Boncz isn’t exactly an Actian employee. Rather, he’s the professor who supervised Marcin Zukowski’s PhD thesis that became Vectorwise, and I chatted with Peter by Skype while he was at home in Amsterdam. I believe his assurances that no Actian personnel sat in on the call.
In other news, Peter is currently working on and optimistic about HyPer. But we literally spent less tana minute talking about that
Before I get to the substance, there’s been a lot of renaming at Actian. To quote Andrew Brust,
… the ParAccel, Pervasive and Vectorwise technologies are being unified under the Actian Analytics Platform brand. Specifically, the ParAccel technology … is being re-branded Actian Matrix; Pervasive’s technologies are rechristened Actian DataFlow and Actian DataConnect; and Vectorwise becomes Actian Vector.
Actian … is now “one company, with one voice and one platform” according to its John Santaferraro
The bolded part of the latter quote is untrue — at least in the ordinary sense of the word “one” — but the rest can presumably be taken as company gospel.
All this is by way of preamble to saying that Peter reached out to me about Actian’s new Vector Hadoop Edition when he blogged about it last June, and we finally talked this week. Highlights include:
- Vectorwise, while being proudly multi-core, was previously single-server. The new Vector Hadoop Edition is the first version with node parallelism.
- Actian’s Vector Hadoop edition uses HDFS (Hadoop Distributed File System) and YARN to manage an Actian-proprietary file format. There is currently no interoperability whereby Hadoop jobs can read these files. However …
- … Actian’s Vector Hadoop edition relies on Hadoop for cluster management, workload management and so on.
- Peter thinks there are two paying customers, both too recent to be in production, who between then paid what I’d call a remarkable amount of money.*
- Roadmap futures* include:
- Being able to update and indeed trickle-update data. Peter is very proud of Vectorwise’s Positional Delta Tree updating.
- Some elasticity they’re proud of, both in terms of nodes (generally limited to the replication factor of 3) and cores (not so limited).
- Better interoperability with Hadoop.
Actian actually bundles Vector Hadoop Edition with DataFlow — the old Pervasive DataRush — into what it calls “Actian Analytics Platform – Hadoop SQL Edition”. DataFlow/DataRush has been working over Hadoop since the latter part of 2012, based on a visit with my then clients at Pervasive that December.
*Peter gave me details about revenue, pipeline, roadmap timetables etc. that I’m redacting in case Actian wouldn’t like them shared. I should say that the timetable for some — not all — of the roadmap items was quite near-term; however, pay no attention to any phrasing in Peter’s blog post that suggests the roadmap features are already shipping.
The Actian Vector Hadoop Edition optimizer and query-planning story goes something like this:
- Vectorwise started with the open-source Ingres optimizer. After a query is optimized, it is rewritten to reflect Vectorwise’s columnar architecture. Peter notes that these rewrites rarely change operator ordering; they just add column-specific optimizations, whatever that means.
- Now there are rewrites for parallelism as well.
- These rewrites all seem to be heuristic/rule-based rather than cost-based.
- Once Vectorwise became part of the Ingres company (later renamed to Actian), they had help from Ingres engineers, who helped them modify the base optimizer so that it wasn’t just the “stock” Ingres one.
As with most modern MPP (Massively Parallel Processing) analytic RDBMS, there doesn’t seem to be any concept of a head-node to which intermediate results need to be shipped. This is good, because head nodes in early MPP analytic RDBMS were dreadful bottlenecks.
Peter and I also talked a bit about SQL-oriented HDFS file formats, such as Parquet and ORC. He doesn’t like their lack of support for columnar compression. Further, in Parquet there seems to be a requirement to read the whole file, to an extent that interferes with Vectorwise’s form of data skipping, which it calls “min-max indexing”.
Frankly, I don’t think the architectural choice “uses Hadoop for workload management and administration” provides a lot of customer benefit in this case. Given that, I don’t know that the world needs another immature MPP analytic RDBMS. I also note with concern that Actian has two different MPP analytic RDBMS products. Still, Vectorwise and indeed all the stuff that comes out Martin Kersten and Peter’s group in Amsterdam has always been interesting technology. So the Actian Vector Hadoop Edition might be worth taking a look at before you redirect your attention to products with more convincing track records and futures.
My client Teradata bought my (former) clients Revelytix and Hadapt.* Obviously, I’m in confidentiality up to my eyeballs. That said — Teradata truly doesn’t know what it’s going to do with those acquisitions yet. Indeed, the acquisitions are too new for Teradata to have fully reviewed the code and so on, let alone made strategic decisions informed by that review. So while this is just a guess, I conjecture Teradata won’t say anything concrete until at least September, although I do expect some kind of stated direction in time for its October user conference.
*I love my business, but it does have one distressing aspect, namely the combination of subscription pricing and customer churn. When your customers transform really quickly, or even go out of existence, so sometimes does their reliance on you.
I’ve written extensively about Hadapt, but to review:
- The HadoopDB project was started by Dan Abadi and two grad students.
- HadoopDB tied a bunch of PostgreSQL instances together with Hadoop MapReduce. Lab benchmarks suggested it was more performant than the coyly named DBx (where x=2), but not necessarily competitive with top analytic RDBMS.
- Hadapt was formed to commercialize HadoopDB.
- After some fits and starts, Hadapt was a Cambridge-based company. Former Vertica CEO Chris Lynch invested even before he was a VC, and became an active chairman. Not coincidentally, Hadapt had a bunch of Vertica folks.
- Hadapt decided to stick with row-based PostgreSQL, Dan Abadi’s previous columnar enthusiasm notwithstanding. Not coincidentally, Hadapt’s performance never blew anyone away.
- Especially after the announcement of Cloudera Impala, Hadapt’s SQL-on-Hadoop positioning didn’t work out. Indeed, Hadapt laid off most or all of its sales and marketing folks. Hadapt pivoted to emphasize its schema-on-need story.
- Chris Lynch, who generally seems to think that IT vendors are created to be sold, shopped Hadapt aggressively.
As for what Teradata should do with Hadapt:
- My initial thought for Hadapt was to just double down, pushing the technology forward, presumably including a columnar option such as the one Citus Data developed.
- But upon reflection, if it made technical sense to merge the Aster and Hadapt products, that would be better yet.
I herewith apologize to Aster co-founder and Hadapt skeptic Tasso Argyros (who by the way has moved on from Teradata) for even suggesting such heresy.
Complicating the story further:
- Impala lets you treat data in HDFS (Hadoop Distributed File System) as if it were in a SQL DBMS. So does Teradata SQL-H. But Hadapt makes you decide whether the data is in HDFS or the SQL DBMS, and it can’t be in both at once. Edit: Actually, see Dan Abadi’s comments below.
- Impala and Oracle’s new SQL-H competitor have daemons running on every data node. So does one option in Hadapt. But I don’t think SQL-H does that yet.
I was less involved with Revelytix that with Hadapt (although I’m told I served as the “catalyst” for the original Teradata/Revelytix partnership). That said, Teradata — like Oracle — is always building out a data integration suite to cover a limited universe of data stores. And Revelytix’ dataset management technology is a nice piece toward an integrated data catalog.
A significant fraction of IT professional services industry revenue comes from data integration. But as a software business, data integration has been more problematic. Informatica, the largest independent data integration software vendor, does $1 billion in revenue. INFA’s enterprise value (market capitalization after adjusting for cash and debt) is $3 billion, which puts it way short of other category leaders such as VMware, and even sits behind Tableau.* When I talk with data integration startups, I ask questions such as “What fraction of Informatica’s revenue are you shooting for?” and, as a follow-up, “Why would that be grounds for excitement?”
*If you believe that Splunk is a data integration company, that changes these observations only a little.
On the other hand, several successful software categories have, at particular points in their history, been focused on data integration. One of the major benefits of 1990s business intelligence was “Combines data from multiple sources on the same screen” and, in some cases, even “Joins data from multiple sources in a single view”. The last few years before application servers were commoditized, data integration was one of their chief benefits. Data warehousing and Hadoop both of course have a “collect all your data in one place” part to their stories — which I call data mustering — and Hadoop is a data transformation tool as well.
And it’s not as if successful data integration companies have no value. IBM bought a few EAI (Enterprise Application Integration) companies, plus top Informatica competitor Ascential, plus Cast Iron Systems. DataDirect (I mean the ODBC/JDBC guys, not the storage ones) has been a decent little business through various name changes and ownerships (independent under a couple of names, then Intersolv/Merant, then independent again, then Progress Software). Master data management (MDM) and data cleaning have had some passable exits. Talend raised $40 million last December, which is a nice accomplishment if you’re French.
I can explain much of this in seven words: Data integration is both important and fragmented. The “important” part is self-evident; I gave examples of “fragmented” a couple years back. Beyond that, I’d say:
- A new class of “engine” can be a nice business — consider for example Informatica/Ascential/Ab Initio, or the MDM players (who sold out to bigger ETL companies), or Splunk. Indeed, much early Hadoop adoption was for its capabilities as a data transformation engine.
- Data transformation is a better business to enter than data movement. Differentiated value in data movement comes in areas such as performance, reliability and maturity, where established players have major advantages. But differentiated value in data transformation can come from “intelligence”, which is easier to excel in as a start-up.
- “Transparent connectivity” is a tough business. It is hard to offer true transparency, with minimal performance overhead, among enough different systems for anybody to much care. And without that you’re probably offering a low-value/niche capability. Migration aids are not an exception; the value in those is captured by the vendor of what’s being migrated to, not by the vendor who actually does the transparent translation. Indeed …
- … I can’t think of a single case in which migration support was a big software business. (Services are a whole other story.) Perhaps Cast Iron Systems came closest, but I’m not sure I’d categorize it as either “migration support” or “big”.
And I’ll stop there, because I’m not as conversant with some of the new “smart data transformation” companies as I’d like to be.
- DBMS transparency layers never seem to sell well (April, 2009)
- ClearStory’s approach to data integration (September, 2013)
- Judging opportunities (July, 2014)
Oracle is announcing today what it’s calling “Oracle Big Data SQL”. As usual, I haven’t been briefed, but highlights seem to include:
- Oracle Big Data SQL is basically data federation using the External Tables capability of the Oracle DBMS.
- Unlike independent products — e.g. Cirro — Oracle Big Data SQL federates SQL queries only across Oracle offerings, such as the Oracle DBMS, the Oracle NoSQL offering, or Oracle’s Cloudera-based Hadoop appliance.
- Also unlike independent products, Oracle Big Data SQL is claimed to be compatible with Oracle’s usual security model and SQL dialect.
- At least when it talks to Hadoop, Oracle Big Data SQL exploits predicate pushdown to reduce network traffic.
And by the way – Oracle Big Data SQL is NOT “SQL-on-Hadoop” as that term is commonly construed, unless the complete Oracle DBMS is running on every node of a Hadoop cluster.
Predicate pushdown is actually a simple concept:
- If you issue a query in one place to run against a lot of data that’s in another place, you could spawn a lot of network traffic, which could be slow and costly. However …
- … if you can “push down” parts of the query to where the data is stored, and thus filter out most of the data, then you can greatly reduce network traffic.
“Predicate pushdown” gets its name from the fact that portions of SQL statements, specifically ones that filter data, are properly referred to as predicates. They earn that name because predicates in mathematical logic and clauses in SQL are the same kind of thing — statements that, upon evaluation, can be TRUE or FALSE for different values of variables or data.
The most famous example of predicate pushdown is Oracle Exadata, with the story there being:
- Oracle’s shared-everything architecture created a huge I/O bottleneck when querying large amounts of data, making Oracle inappropriate for very large data warehouses.
- Oracle Exadata added a second tier of servers each tied to a subset of the overall storage; certain predicates are pushed down to that tier.
- The I/O between Exadata’s two sets of servers is now tolerable, and so Oracle is now often competitive in the high-end data warehousing market,
Oracle evidently calls this “SmartScan”, and says Oracle Big Data SQL does something similar with predicate pushdown into Hadoop.
Oracle also hints at using predicate pushdown to do non-tabular operations on the non-relational systems, rather than shoehorning operations on multi-structured data into the Oracle DBMS, but my details on that are sparse.
- Chris Kanaracus’ coverage of the announcement quotes me at length.
As part of my series on the keys to and likelihood of success, I outlined some examples from the DBMS industry. The list turned out too long for a single post, so I split it up by millennia. The part on 20th Century DBMS success and failure went up Friday; in this one I’ll cover more recent events, organized in line with the original overview post. Categories addressed will include analytic RDBMS (including data warehouse appliances), NoSQL/non-SQL short-request DBMS, MySQL, PostgreSQL, NewSQL and Hadoop.
DBMS rarely have trouble with the criterion “Is there an identifiable buying process?” If an enterprise is doing application development projects, a DBMS is generally chosen for each one. And so the organization will generally have a process in place for buying DBMS, or accepting them for free. Central IT, departments, and — at least in the case of free open source stuff — developers all commonly have the capacity for DBMS acquisition.
In particular, at many enterprises either departments have the ability to buy their own analytic technology, or else IT will willingly buy and administer things for a single department. This dynamic fueled much of the early rise of analytic RDBMS.
Buyer inertia is a greater concern.
- A significant minority of enterprises are highly committed to their enterprise DBMS standards.
- Another significant minority aren’t quite as committed, but set pretty high bars for new DBMS products to cross nonetheless.
- FUD (Fear, Uncertainty and Doubt) about new DBMS is often justifiable, about stability and consistent performance alike.
A particularly complex version of this dynamic has played out in the market for analytic RDBMS/appliances.
- First the newer products (from Netezza onwards) were sold to organizations who knew they wanted great performance or price/performance.
- Then it became more about selling “business value” to organizations who needed more convincing about the benefits of great price/performance.
- Then the behemoth vendors became more competitive, as Teradata introduced lower-price models, Oracle introduced Exadata, Sybase got more aggressive with Sybase IQ, IBM bought Netezza, EMC bought Greenplum, HP bought Vertica and so on. It is now hard for a non-behemoth analytic RDBMS vendor to make headway at large enterprise accounts.
- Meanwhile, Hadoop has emerged as serious competitor for at least some analytic data management, especially but not only at internet companies.
Otherwise I’d say:
- At large enterprises, their internet operations perhaps excepted:
- Short-request/general-purpose SQL alternatives to the behemoths — e.g. MySQL, PostgreSQL, NewSQL — have had tremendous difficulty getting established. The last big success was the rise of Microsoft SQL Server in the 1990s. That’s why I haven’t mentioned the term mid-range DBMS in years.
- NoSQL/non-SQL has penetrated large enterprises mainly for a few specific use cases, for example the lists I posted for MongoDB or graph databases.
- Internet-only companies have few inertia issues when it comes to database managers. They’ll consider anything they regard as being in their price ballpark (which is however often restricted to open source). I think part of the reason is that as quickly as they rewrite their applications, DBMS are vastly less “strategic” to them than they are to most larger enterprises.
- The internet operations of large companies — especially large retailers — in many cases behave like internet-only companies, but in many other cases behave like the rest of the enterprise.
The major reasons for DBMS categories to get established in the first place are:
- Performance and/or scalability (many examples).
- Developer features (for example dynamic schema).
- License/maintenance cost (for example several open source categories).
- Ease of installation and administration (for example open source again, and also data warehouse appliances).
Those same characteristics are major bases for competition among members of a new category, although as noted above behemoth-loyalty can also come into play.
Cool-vs.-weird tradeoffs are somewhat secondary among SQL DBMS.
- There’s not much of a “cool” factor, because new products aren’t that different in what they do vs. older ones.
- There’s not a terrible “weird” factor either, but of course any smaller offering faces FUD, and also …
- … appliances are anti-strategic for many buyers, especially ones who demand a smooth path to the cloud.)
They’re huge, however, in the non-SQL world. Most non-SQL data managers have a major “weird” factor. Fortunately, NoSQL and Hadoop both have huge “cool” cred to offset it. XML/XQuery unfortunately did not.
Finally, in most DBMS categories there are massive issues with product completeness, more in the area of maturity than that of whole product. The biggest whole product issues are concentrated on the matter of interoperating with other software — business intelligence tools, packaged applications (if relevant to the category), etc. Most notably, the handful of DBMS that are certified to run SAP share a huge market that other DBMS can’t touch. But BI tools are less of a differentiator — I yawn when vendors tell me they are certified for/partnered with MicroStrategy, Tableau, Pentaho and Jaspersoft, and I’m surprised at any product that isn’t.
DBMS maturity has a lot of aspects, but the toughest challenges are concentrated in two main areas:
- Reliability, especially but not only in short-request use cases.
- Performance across a great variety of use cases. I observe frequently that performance in best-case scenarios, performance in the lab and performance in real-world environments are much further apart than vendors like to think.
- Maturity demands seem to be much higher for SQL DBMS than for NoSQL.
- I think this is one of several reasons NoSQL has been much more successful than NewSQL.
- It’s why I think MarkLogic’s “Enterprise NoSQL” positioning is a mistake.
- As for MySQL:
- MySQL wasn’t close to reliable enough for enterprises to trust it until InnoDB became the default storage engine.
- MySQL 5 point releases have added major features, or decent performance for major features. I’ll confess to having lost track of what’s been fixed and what’s still missing.
- In saying all that I’m holding MySQL to a much higher maturity standard than I’m holding NoSQL — because that’s what I think enterprise customers do.
- PostgreSQL “should” be doing a lot better than it is. I have an extremely low opinion of its promoters, and not just for personal reasons. (That said, the personal reasons don’t just apply to EnterpriseDB anymore. I’ve also run out of patience waiting for Josh Berkus to retract untruths he posted about me years ago.)
- SAP HANA checks boxes for performance (In-memory rah rah rah!!) and whole product (Runs SAP!!). That puts it well ahead of most other newish SQL DBMS, purely analytic ones perhaps excepted.
- Any other new short-request SQL DBMS that sounds like is has traction is also memory-centric.
- Analytic RDBMS are in most respects held to lower maturity standards than DBMS used for write-intensive workloads. Even so, products in the category are still frequently tripped up by considerations of concurrent performance and mixed workload management.
There have been 1,470 previous posts in the 9-year history of this blog, many of which could serve as background material for this one. A couple that seem particularly germane and didn’t get already get linked above are:
I’m commonly asked to assess vendor claims of the kind:
- “Our system lets you do multiple kinds of processing against one database.”
- “Otherwise you’d need two or more data managers to get the job done, which would be a catastrophe of unthinkable proportion.”
So I thought it might be useful to quickly review some of the many ways organizations put multiple data stores to work. As usual, my bottom line is:
- The most extreme vendor marketing claims are false.
- There are many different choices that make sense in at least some use cases each.
Horses for courses
It’s now widely accepted that different data managers are better for different use cases, based on distinctions such as:
- Short-request vs. analytic.
- SQL vs. non-SQL (NoSQL or otherwise).
- Expensive/heavy-duty vs. cheap/easy-to-support.
Vendors are part of this consensus; already in 2005 I observed
For all practical purposes, there are no DBMS vendors left advocating single-server strategies.
Vendor agreement has become even stronger in the interim, as evidenced by Oracle/MySQL, IBM/Netezza, Oracle’s NoSQL dabblings, and various companies’ Hadoop offerings.
Multiple data stores for a single application
We commonly think of one data manager managing one or more databases, each in support of one or more applications. But the other way around works too; it’s normal for a single application to invoke multiple data stores. Indeed, all but the strictest relational bigots would likely agree:
- It’s common and sensible to manage authentication and authorization data in its own data store. Commonly, the data format is LDAP (Lightweight Directory Access Protocol).
- It’s common and sensible to manage the “content” and “e-commerce transaction records” aspects of websites separately.
- Even beyond that case, there are often performance reasons to manage BLOBs (Binary Large OBjects) outside your relational database.
- Internet “interaction” data is also often best managed outside an RDBMS, in part because of its very non-tabular data structures.
The spectacular 2010 JP Morgan Chase outage was largely caused, I believe, by disregard of these precepts.
There also are cases in which applications dutifully get all their data via SQL queries, but send those queries to two or more DBMS. Teradata is proud that its systems can support rather transactional queries (for example in call-center use cases), but the same application may read from and write to a true OTLP database as well.
Further, many OLTP (OnLine Transaction Processing) applications do some fraction of their work via inbound or outbound messaging. Many buzzwords can come into play here, including but not limited to:
- SOA (Service-Oriented Architecture). This is the most current and flexible one.
- EAI (Enterprise Application Integration). This was a hot concept in the late 1990s, but was generally implemented with difficulties that SOA was later designed to alleviate.
- Message-oriented middleware (MOM) and Publish/Subscribe. These are even older, and overlap greatly.
Finally, every dashboard that combines information from different data stores could be assigned to this category as well.
Multiple storage approaches in a single DBMS
In theory, a single DBMS could operate like two or more different ones glued together. A few functions should or must be centralized, such as administration, and communication with the outside world (connection handling, parsing, etc.). But data storage, query execution and so on could for the most part be performed by rather loosely coupled subsystems. And so you might have the best of both worlds — something that’s multiple data stores in the ways you want that diversity, but a single system in how it fits into your environment.
I discussed this idea last year with cautious optimism, writing:
So will these trends succeed? The forgoing caveats notwithstanding, my answers are more Yes than No.
- … multi-purpose DBMS will likely always have performance penalties, but over time the penalties should become small enough to be affordable in most cases.
- Machine-generated data and “content” both call for multi-datatype DBMS. And taken together, those are a large fraction of the future of computing. Consequently …
- … strong support for multiple datatypes and DMLs is a must for “general-purpose” RDBMS. Oracle and IBM [have] been working on that for 20 years already, with mixed success. I doubt they’ll get much further without a thorough rewrite, but rewrites happen; one of these decades they’re apt to get it right.
In 2005 I had been more ambivalent, in part because my model was a full 1990s-dream “universal” DBMS:
IBM, Oracle, and Microsoft have all worked out ways to have integrated query parsing and query optimization, while letting storage be more or less separate. More precisely, Oracle actually still sticks everything into one data store (hence the lack of native XML support), but allows near-infinite flexibility in how it is accessed. Microsoft has already had separate servers for tabular data, text, and MOLAP, although like Sybase, it doesn’t have general datatype extensibility that it can expose to customers, or exploit itself to provide a great variety of datatypes. IBM has had Oracle-like extensibility all along, although it hasn’t been quite as aggressive at exploiting it; now it’s introduced a separate-server option for XML.
That covers most of the waterfront, but I’d like to more explicitly acknowledge three trends:
- Among other things, Hadoop is a collection of DBMS (HBase, Impala, et al.) that in some cases are very loosely coupled to each other. The question is less how well the various data stores work together, and more how mature any one of them is on its own.
- The multiple-data-models idea has been extended into schema-on-need, which is sometimes but not always housed in Hadoop.
- Even on the relational side, multiple storage capabilities exist in one product.
- Vertica was designed that way from the get-go. (Like the old joke about police duos, one is to read and one is to write.)
- IBM, Microsoft and Oracle have all recently added some kind of in-memory columnar capability.
- Teradata, Aster (before Teradata bought them), Greenplum and Vertica all added some variant on row/column dual stores.
The pessimist thinks the glass is half-empty.
The optimist thinks the glass is half-full.
The engineer thinks the glass was poorly designed.
Most of what I wrote in Part 1 of this post was already true 15 years ago. But much gets added in the modern era, considering that:
- Clusters will have node hiccups more often than single nodes will. (Duh.)
- Networks are relatively slow even when uncongested, and furthermore congest unpredictably.
- In many applications, it’s OK to sacrifice even basic-seeming database functionality.
And so there’s been innovation in numerous cluster-related subjects, two of which are:
- Distributed query and update. When a database is distributed among many modes, how does a request access multiple nodes at once?
- Fault-tolerance in long-running jobs.When a job is expected to run on many nodes for a long time, how can it deal with failures or slowdowns, other than through the distressing alternatives:
- Start over from the beginning?
- Keep (a lot of) the whole cluster’s resources tied up, waiting for things to be set right?
Distributed database consistency
When a distributed database lives up to the same consistency standards as a single-node one, distributed query is straightforward. Performance may be an issue, however, which is why we have seen a lot of:
- Analytic RDBMS innovation.
- Short-request applications designed to avoid distributed joins.
- Short-request clustered RDBMS that don’t allow fully-general distributed joins in the first place.
But in workloads with low-latency writes, living up to those standards is hard. The 1980s approach to distributed writing was two-phase commit (2PC), which may be summarized as:
- A write is planned and parceled out to occur on all the different nodes where the data needs to be placed.
- Each node decides it’s ready to commit the write.
- Each node informs the others of its readiness.
- Each node actually commits.
Unfortunately, if any of the various messages in the 2PC process is delayed, so is the write. This creates way too much likelihood of work being blocked. And so modern approaches to distributed data writing are more … well, if I may repurpose the famous Facebook slogan, they tend to be along the lines of “Move fast and break things”,* with varying tradeoffs among consistency, other accuracy, reliability, functionality, manageability, and performance.
By the way — Facebook recently renounced that motto, in favor of “Move fast with stable infrastructure.” Hmm …
Back in 2010, I wrote about various approaches to consistency, with the punch line being:
A conventional relational DBMS will almost always feature RYW consistency. Some NoSQL systems feature tunable consistency, in which — depending on your settings — RYW consistency may or may not be assured.
The core ideas of RYW consistency, as implemented in various NoSQL systems, are:
- Let N = the number of copies of each record distributed across nodes of a parallel system.
- Let W = the number of nodes that must successfully acknowledge a write for it to be successfully committed. By definition, W <= N.
- Let R = the number of nodes that must send back the same value of a unit of data for it to be accepted as read by the system. By definition, R <= N.
- The greater N-R and N-W are, the more node or network failures you can typically tolerate without blocking work.
- As long as R + W > N, you are assured of RYW consistency.
That bolded part is the key point, and I suggest that you stop and convince yourself of it before reading further.
Eventually , Dan Abadi claimed that the key distinction is synchronous/asynchronous — is anything blocked while waiting for acknowledgements? From many people, that would simply be an argument for optimistic locking, in which all writes go through, and conflicts — of the sort that locks are designed to prevent — cause them to be rolled back after-the-fact. But Dan isn’t most people, so I’m not sure — especially since the first time I met Dan was to discuss VoltDB predecessor H-Store, which favors application designs that avoid distributed transactions in the first place.
One idea that’s recently gained popularity is a kind of semi-synchronicity. Writes are acknowledged as soon as they arrive at a remote node (that’s the synchronous part). Each node then updates local permanent storage on its own, with no further confirmation. I first heard about this in the context of replication, and generally it seems designed for replication-oriented scenarios.
Finally, let’s consider fault-tolerance within a single long-running job, whether that’s a big query or some other kind of analytic task. In most systems, if there’s a failure partway through a job, they just say “Oops!” and start it over again. And in non-extreme cases, that strategy is often good enough.
Still, there are a lot of extreme workloads these days, so it’s nice to absorb a partial failure without entirely starting over.
- Hadoop MapReduce, which stores intermediate results anyway, finds it easy to replay just the parts of the job that went awry.
- Spark, which is more flexible in execution graph and data structures alike, has a similar capability.
Additionally, both Hadoop and Spark support speculative execution, in which several clones of a processing step are executed at once (presumably on different nodes), to hedge against the risk that any one copy of the process runs slowly or fails outright. According to my notes, speculative execution is a major part of NuoDB’ architecture as well.
I’ve rambled on for two long posts, which seems like plenty — but this survey is in no way complete. Other subjects I could have covered include but are hardly limited to:
- Occasionally-connected operation, which for example is a design point of CouchDB, SQL Anywhere (sort of), and most kinds of mobile business intelligence.
- Avoiding planned downtime — i.e., operating despite self-inflicted wounds.
- Data cleaning and master data management, both of which exist in large part to fix errors people have made in the past.
Writing data management or analysis software is hard. This post and its sequel are about some of the reasons why.
When systems work as intended, writing and reading data is easy. Much of what’s hard about data management is dealing with the possibility — really the inevitability — of failure. So it might be interesting to survey some of the many ways that considerations of failure come into play. Some have been major parts of IT for decades; others, if not new, are at least newly popular in this cluster-oriented, RAM-crazy era. In this post I’ll focus on topics that apply to single-node systems; in the sequel I’ll emphasize topics that are clustering-specific.
Major areas of failure-aware design — and these overlap greatly — include:
- Backup and restore. In its simplest form, this is very basic stuff. That said — any decent database management system should let backups be made without blocking ongoing database operation, with the least performance impact possible.
- Logging, rollback and replay. Logs are essential to DBMS. And since they’re both ubiquitous and high-performance, logs are being used in ever more ways.
- Locking, latching, transactions and consistency. Database consistency used to be enforced in stern and pessimistic ways. That’s changing, big-time, in large part because of the requirements of …
- … distributed database operations. Increasingly, modern distributed database systems are taking the approach of getting work done first, then cleaning up messes when they occur.
- Redundancy and replication. Parallel computing creates both a need and an opportunity to maintain multiple replicas of data at once, in very different ways than the redundancy and replication of the past.
- Fault-tolerant execution. When one node is inoperative, inaccessible, overloaded or just slow, you may not want a whole long multi-node job to start over. A variety of techniques address this need.
In a single-server, disk-based configuration, techniques for database fault-tolerance start:
- Database changes (inserts, deletes, updates) are applied expeditiously to disk. Furthermore …
- … a log is kept of the (instructions for) changes. If a change is detected as not going through, it can be reapplied.
- Data is often kept in multiple copies automagically, whether that is governed by the storage systems or the DBMS itself, with background resyncing if one copy is known to go bad.
- These ideas can be pushed further into:
- High availability — the database is mirrored onto storage that is controlled by a second server, which runs the same software as the primary.
- Disaster recovery — same idea as HA, but off-site to protect against site disasters. DR often cuts various corners, as compared to same-site HA, in the speed and/or quality of service with which the remote instance would take over from the primary site.
Valuable though they are, none of these techniques protects against data corruption that occurs via software errors or security breaches, because in those cases the system will likely do a great job of making incorrect changes consistently across all copies of the data. Hence there also needs to be a backup/restore capability, key aspects of which start:
- You periodically create and lock down snapshots of the database.
- In case of comprehensive failure, you load the most recent snapshot you trust, and roll forward by re-applying changes memorialized in the log. (Of course, you avoid re-applying spurious changes that should not have occurred.)
For many users, it is essential that backups be online/continuous, rather than requiring the database to periodically be taken down.
None of this is restricted to what in a relational database would be single-row or at least single-table changes. Transaction semantics cover the case that several changes must either all change or all fail together. Typically all the changes are made; the system observes that they’ve all been made; only then is a commit issued which makes the changes stick. One complexity of this approach is that you need a way to quickly undo changes that don’t get committed.
Locks, latches and timestamps
If a database has more than one user, two worrisome possibilities arise:
- Two conflicting changes might be attempted against the same data.
- One user might try to read data at the moment another user’s request was changing it.
So it is common to lock the portion of the database that is being changed. A major area of competition in the early 1990s — and a big contributor to Sybase’s decline — was the granularity of locks (finer-grained is better, with row-level locking being the relational DBMS gold standard).
In-memory locks are generally called latches; I don’t know what the other differences between locks and latches are.
Increasingly many DBMS are designed with an alternative approach, MVCC (Multi-Version Concurrency Control); indeed, I’m now startled when I encounter one that makes a different choice. The essence of MVCC is that, to each portion (e.g. row) of data, the system appends very granular metadata — a flag for validation/invalidation and/or a timestamp. Except for the flags, the data is never changed until a cleanup operation; rather, new data is written, with later timestamps and validating flags. The payoff is that, while it may still be necessary to lock data against the possibility of two essentially simultaneous writes, reads can proceed without locks, at the cost of a minuscule degree of freshness. For if an update is in progress that affects the data needed for a read, the read just targets versions of the data slightly earlier in time.
MVCC has various performance implications — writes can be faster because they are append-mainly, reads may be slower, and of course the cleanup is needed. These tradeoffs seem to work well even for the query-intensive workloads of analytic RDBMS, perhaps because MVCC fits well with their highly sequential approach to I/O.
A DBMS that truly ran only in RAM would lose data each time the power turned off. Hence a memory-centric DBMS will usually also have some strategy for persisting data.