The graph below displays some major players in healthcare big data. Note Aerospike’s position at the top of the inner circle.

“Make note of the names in the inner circle. They’re the pure plays with the newest science—and are likely to get gobbled up by the growth-hungry incumbents on the outside.” – Forbes.com

See the full Industry Atlas graph on Forbes.com.

In a recent interview with Roberto Zicari, the editor of ODBMS Industry Watch blog, Aerospike co-founder and CTO Brian Bulkowski described a few use cases in which Aerospike’s sub-millisecond response times and 100% uptime can put companies in these industries ahead of their competitors in terms of reliability and user satisfaction.

“Other use cases include massive multi-player gaming, where a shared game state (at scale) is key to a great user experience, and retaining peak capacity is critical.”

“As a row-based store with very high levels of read and write concurrency, we perform well where behavior and interactions are concerned – whether those interactions are user-oriented, such as advertising and gaming; or machine to machine, such as capital markets or the emerging Internet of Things.”

Read the rest of the interview here.

Mountain View, CA – May 20, 2013 –The big data discussion is shifting to “fast data” as more companies seek to harness the unprecedented volumes of information being generated at high velocity via Web, social and mobile interactions and sensors. These businesses are rolling out new applications that respond immediately to what people, devices and systems are doing now. In turn, this is driving the demand for a new class of highly distributed in-memory databases that also are optimized for fast, stable flash-based storage.

In the wake of this fast data wave, Aerospike, Inc. has attracted in-memory database pioneer Marie-Anne Neimat to join the real-time database company as an advisor. Ms. Neimat co-founded TimesTen, Inc. the first company to develop and commercialize an in-memory relational database. Most recently, she served as vice president of engineering for three of Oracle’s databases, including TimesTen, which Oracle acquired. Bringing a track record of database innovation and deep insight into CIOs’ demands, she will play a strategic role in helping Aerospike to build on its technology leadership and rapid market adoption.

The Future of Fast Data in Production Today
Already, customers have the enterprise-grade Aerospike real-time hybrid DRAM and flash memory database in production supporting fast data applications within some of the world’s most data-intensive environments. These companies’ real-time big data platforms rely on Aerospike to consistently respond within milliseconds, even while scaling to hundreds of thousands of transactions per second (TPS) and managing billions of objects and terabytes of data—all with 100% uptime.

“The ubiquity of smart devices is heralding a new era of fast data where the value of applications lies in the ability to sense and respond to Internet-scale data in real time. However, it is also challenging databases to cost-effectively and reliably deliver speed at scale,” Ms. Neimat observed. “Aerospike is at the forefront of the database revolution in addressing this challenge with a unique hybrid memory approach and scale-up/scale-out architecture that is deployed at dozens of Internet-scale sites today. I’m excited to join the Aerospike team in building on this tremendous momentum.”

Joining Forces With In-Memory Database Innovator
Marie-Anne Neimat was at Oracle Corporation where she most recently served as vice president of development for three database products: Oracle TimesTen In-Memory Database, Oracle Berkeley Database, and Oracle NoSQL Database. Prior to joining Oracle, Ms. Neimat was co-founder of TimesTen, Inc. the first company to develop and commercialize an in-memory relational database. There, she served as vice president of engineering and board member, and in 2004 she was recognized on the Fast Company Fast 50 list of ordinary people doing extraordinary things. TimesTen, Inc. was acquired by Oracle in 2005. Prior to founding her own company, Ms. Neimat worked at HP Labs where she managed several database research projects. Ms. Neimat holds a BSC in mathematics from Stanford University and a PhD in computer science from UC Berkeley. She holds several patents and is the author of many publications in refereed conferences and journals.

“It is a deep honor to have Marie-Anne join us as an advisor. With her roles in leading in-memory database design at Oracle and TimesTen, she brings valuable insights into the data demands facing CTOs today, and a track record of success in growing database adoption,” said Bruce Fram, Aerospike CEO. “With Marie-Anne joining Don Haderle, father of IBM DB2, and Roger Sippl, Informix founder and Silicon Valley pioneer, we have an incredible advisory board with an unmatched breadth and depth of experience. These advisors, along with the technology rock stars driving our database development, are solidly positioning us to lead the fast data revolution being fueled by the Internet economy.”

About Aerospike
Aerospike, Inc. offers the only real-time Internet-scale database that delivers predictable high performance for mission-critical applications. Aerospike’s hybrid memory (DRAM and native flash) architecture scales up and out, consistently processing over 500k transactions per second per node with sub-millisecond latency. With automatic fail-over, replication, and cross data center synchronization, the Aerospike database reliably stores billions of objects and terabytes of data—while providing 100% uptime and 17x better TCO than other NoSQL databases. Customers accelerating their business with Aerospike include adMarketplace, Bluekai, Chango, eXelate, Federated Media, The Trade Desk and [x +1].

Mountain View, CA – May 13, 2013 – Aerospike, provider of the real-time Internet-scale database, today announced that it has been named one of the Lightning Round winners for the Big Data track of TiEcon 2013.  Lightning Round winners are recognized for their leadership among their peers and their game-changing approaches and technologies that are likely to disrupt existing markets and entrenched players. Aerospike and other Lightning Round winners will be honored at TiEcon 2013 on Friday, May 17, at the Santa Clara Convention Center.

“We realized that Internet-scale consumer services would lead the big data-driven application revolution. So we built our NoSQL database for speed at scale, with hybrid DRAM and flash memory and maximum reliability through self-managing, automated cluster technology,” said Srini V. Srinivasan, Aerospike founder and vice president of engineering and operations. “We are honored to be named a Lightning Round winner at the prestigious TiEcon event. It recognizes our groundbreaking approach and proven success in mission-critical Internet-scale deployments.”

“To honor the startups that are showing signs of early success and huge potential for success, TiEcon has created a new category of Lightning Round winners who will have the chance to pitch their stories to the full Big Data track, comprising of investors, industry executives, prospects, analysts and media,” said R. Paul Singh, co-convener of TiEcon 2013.

Aerospike is a powerful, mission-critical NoSQL database for real-time big data applications. The Aerospike Hybrid (DRAM and Flash) Memory System™, Aerospike Smart Clusters™ and Aerospike Cross Data Center Replication (XDR) capabilities implement an architecture that is ACID, in production for more than three years with zero downtime, processing tens of terabytes of data and 30 billion transactions per day with 2-3 millisecond reads/writes while meeting a 99.9% service-level agreement (SLA).

TiEcon is TiE’s premiere annual event and is the largest conference for entrepreneurs in the world. TiEcon 2013 offers full day tracks on three of the hottest industry verticals of Big Data, Mobility and Software Defined Infrastructure on May 17^th, 2013. The second day of the conference, May 18^th, focuses on issues of entrepreneurship including how-tos, Women, Youth, Professional Services, Global and Social Entrepreneurship.

For more information on TiEcon 2013, visit www.tiecon.org.

About TiEcon
TiEcon is the world’s largest conference for entrepreneurs and intrapreneurs with loyal participation from top technology companies, leading venture capital firms, and global service providers. TiEcon 2012 attracted 3,000+ attendees from across the world—including CEOs of established companies to first-time entrepreneurs creating new companies, to leading investment professionals and corporate executives. TiEcon was listed by Worth Magazine in their September 2011 issue to be among the 10 Best Conferences for Ideas and Entrepreneurship along with TED and the World Economic Forum. For more information on TiEcon, visit www.tiecon.org.

About Aerospike
Aerospike, Inc. offers the only real-time Internet-scale database that delivers predictable high performance for mission-critical applications. Aerospike’s hybrid memory (DRAM and native flash) architecture scales up and out, consistently processing over 500k transactions per second per node with sub-millisecond latency. With automatic fail-over, replication, and cross data center synchronization, the Aerospike database reliably stores billions of objects and terabytes of data—while providing 100% uptime and 17x better TCO than other NoSQL databases. Customers accelerating their business with Aerospike include adMarketplace, Bluekai, Chango, eXelate, Federated Media, The Trade Desk and [x +1].

The Conclusion
The results of the Maximum Throughput tests are represented in the graph below.

“Aerospike was an order of magnitude better than anybody else. And Couchbase didn’t run at all. Couchbase, in theory, the underlying disk was doing about 40,000 operations per second, but it got bottlenecked by RAM.”

“So, I think the conclusions for the performance part are obvious. Aerospike claims to be an SSD optimized database. They’ve written their own drivers to access the raw hardware. That was very clearly true in our tests.”

“What limited these things and caused these latency slowdowns? For the synchronous case, it was disk IOPS (Input/Out Operations per Second). For the asynchronous case… it wasn’t necessarily that easy to know what was going on”.

“What do I mean by synchronous here? This just means the reliable use case. That means we will wait for the replica, in this case, our replication factor was two, so there’s one replica. We will wait for that replica to confirm that it’s been written to before the marking the write as OK. And, if possible, we will confirm that the disk has been written. OK, now, confirming that the disk has been written is not the whole story, because disk could be written, but the fsync might not have run yet, so it might actually be still in, sort of, controller memory, and that differs from database to database. But the replica into the other replica is the critical point, because if you lose the node, you still have the data.”

Figure 1. Vertical axis shows latency in milliseconds and horizontal axis shows execution time in milliseconds.

Figure 2. Vertical axis shows throughput operations per second and horizontal axis shows execution time in milliseconds.

Why does latency matter? Read more.

Mountain View, CA – May 7, 2013 – From Madison Avenue to Moscow, online advertising is emerging as the fastest growing ad sector in the world. In 2012, a burgeoning Russian online advertising market reached nearly $2 billion and is expected to double to $4 billion by 2015. Tinkoff Digital is helping to accelerate Russian ad market growth through the development of new real-time technologies for the country’s booming online and mobile advertising sectors. A new case study examines how Tinkoff has increased performance while cutting operational and hardware costs associated with its previous NoSQL database using the Aerospike real-time hybrid memory and flash database. The full Tinkoff case study is available at: http://www.aerospike.com/why-aerospike/customers.

Tinkoff launched Russia’s first real-time bidding (RTB) platforms for online and mobile advertisers: Data Mind and MADNET. More recently, Tinkoff strengthened its position as an early ad tech leader in Russia with the release of a new data management platform (DMP). Powered by the Aerospike real-time database, Tinkoff’s RTB-optimized DMP responds within milliseconds even as it scales, to collect and manage data linked to hundreds of millions of Russian Internet users.

Tinkoff Case Study
The new case study examines the factors that led Tinkoff to replace its existing NoSQL database with Aerospike. It also discusses how Tinkoff has deployed Aerospike across three data centers in Russia, the United Kingdom and United States to bring data closer to customers while ensuring 100% uptime. Key benefits that Tinkoff has realized with Aerospike include:

• Delivers 20,000 parallel read/update operations per second with less than a 7-millisecond latency on just two servers, while managing hundreds of millions of profiles of Russian Internet users.
• Allows Tinkoff to achieve the same performance as its previous open source database with just a fraction of the hardware through Aerospike’s hybrid architecture combining RAM speed with flash reliability.
• Eliminates the need for workarounds in order to prevent performance degradation during parallel read/update operations, which Tinkoff experienced with its previous NoSQL database.
• Reduces work for engineers through Aerospike’s automated cluster configuration, significantly reducing maintenance and support costs.
• Enables Tinkoff to locate closer to consumers, deliver 100% uptime, and minimize hardware across data centers through Aerospike’s cross data center replication.

 
“Aerospike’s performance combined with the ability to reduce the costs of maintenance, support and hardware make Aerospike a truly attractive data management solution,” said Konstantin Kruglov, Tinkoff Digital CTO. “Given the results of adopting Aerospike for our data management platform, we are now planning to completely replace our previous NoSQL solution with Aerospike databases across our supply-side platform, demand-side platform, and mobile ad network.”

“Tinkoff is a leader in the Russian Internet ad market, which is expected to double in the next three years,” said Brian Bulkowski, Aerospike founder and CTO. “We are thrilled by the company’s success in implementing our Aerospike real-time database, not only to enhance performance and operational efficiency today, but also to provide a platform for sustainable growth that will enable Tinkoff to expand and meet market demand.”

About Tinkoff
Tinkoff Digital is а Russian developer of innovative Internet and mobile advertising technologies. Among its solutions are MADNET, the first RTB platform in Russia for mobile devices, and Data Mind the first RTB platform in Russia for the Internet. Tinkoff Digital is Google’s first partner in Russia providing mobile advertising impressions through Google DoubleClick Ad Exchange. The company has also signed agreements with SUP Media, one of Russia’s biggest online media companies, as well as leading international mobile networks including SMAATO, Mopub, Mobclix and Hunt Mobile Ads. For more information, visit http://tinkoffdigital.ru.

About Aerospike
Aerospike, Inc. offers the only real-time Internet-scale database that delivers predictable high performance for mission-critical applications. Aerospike’s hybrid memory (DRAM and native flash) architecture scales up and out, consistently processing over 500k transactions per second per node with sub-millisecond latency. With automatic fail-over, replication, and cross data center synchronization, the Aerospike database reliably stores billions of objects and terabytes of data—while providing 100% uptime and 17x better TCO than other NoSQL databases. Customers accelerating their business with Aerospike include adMarketplace, Bluekai, Chango, eXelate, Federated Media, The Trade Desk and [x +1].

Change happens over time, but innovation happens in real-time.

At Tinkoff Digital, we are working to make real-time bidding (RTB) technology accessible to Russia’s fast-growing Internet advertising market, which is expected to double from $2 billion in 2012 to $4 billion in 2015.

Companies around the world have been seeking to optimize the return on their advertising spend, and RTB is the most effective way to ensure that every online ad is tailored as specifically as possible to potential customers.

To monetize this opportunity, we built and delivered two of Russia’s first RTB platforms for online and mobile advertisers: Data Mind and MADNET. Data Mind lets advertisers bid in real time for traffic on advertising networks based on either cost per impression (CPM) or cost per click (CPC). MADNET serves as a single access point for all Russian mobile traffic and ensures extremely accurate prediction of performance and automatically optimizes advertising impressions for maximum impact.

With massive volumes of aggregated data pouring in across our platforms and networks, we saw the need to develop a data management platform (DMP) that would act as a central hub for collecting and managing large amounts of consumer information. To stay true to our vision of delivering real-time innovation, the DMP also had to be optimized for RTB.

We initially deployed a popular open-source NoSQL database for six months with satisfactory results. However, we replaced it with Aerospike after tests revealed that Aerospike exceeded our performance demands and would enable us to achieve better operational and financial efficiency than we were realizing with our existing NoSQL solution.

Since deploying Aerospike, the database has continued to enhance our DMP’s responsiveness and availability through Aerospike’s hybrid memory and flash architecture and an elegant approach to cross data center replication. Meanwhile, we have been able to reduce our support, maintenance and hardware costs.

Given the results of adopting Aerospike for our data management platform, we are now planning to completely replace our previous NoSQL solution with Aerospike databases across our supply-side platform, demand-side platform, and mobile ad network.

We recently worked with Aerospike on a new case study, which can be found at: http://www.aerospike.com/why-aerospike/customers.

To set the scene for his audience, Ben outlined why Thumbtack decided to do this benchmark evaluation. As a software development services firm that specializes in building and integrating scalable applications and systems, clients frequently turn to them for advice in choosing a NoSQL database solution.

“One of the things that comes up really often when we talk to clients, one of the first things they ask us is, ‘What NoSQL database should we use?’”

“Some of the problems that people have is … There’s a lot of bad reports out there.”

“Why use NoSQL at all? You need to handle a large number of transactions, you need to handle a lot of data, and you don’t want to deal with crazy problems and tremendous manual overhead when systems go down and you need to recover and so forth.”

The Competition
For this benchmark, Thumbtack selected four of the leading NoSQL databases; Aerospike, Couchbase, Cassandra, and MongoDB. Ben gave the audience an overview of each database’s read and write process so they could see the similarities and differences in how each one functioned.

“Couchbase is, in many ways, the easiest to describe. It’s essentially memcache with replication and persistence, and is the hypothetical cluster of six nodes. There’s data on various shards spread across it, and each one of these, we’re assuming a replication factor of three, here. So, each node is a master for one shard, and there are two slaves for each shard. And Couchbase is a consistent database, but it achieves this very simply. It just always reads and writes from an NAS disk. The fast scenario, very easy to describe, it reads and writes to master and returns immediately, and doesn’t wait for the data to replicate. You have the option of asking Couchbase to wait until that data is replicated. Now you have consistency… Couchbase is always consistent, because it always writes to the master. However, if that master node goes down, you’ve lost data.”

“MongoDB works very much the same way, but it uses a more traditional MySQL master-slave replication model. But essentially, you either wait for the writes to succeed and get propogated, or you don’t.”

“Cassandra enforces consistency in a simple way. When you write, you can either wait for one, or all, or a plurum of the nodes to acknowledge the write. And the same when you read. So, if you read data, if the number of nodes you read from plus the number of nodes you write from is more than your replication factor, you’re going to get consistent data. So, that might mean you write to all the nodes every time, and read from any one. It might mean you write to one, but you always read from all of them, and just take the most recent, or you write to two and you read from two, and take the most recent. That’s all built into Cassandra.”

“Aerospike is simpler to describe. It works just like Cassandra when running in fast mode for its writes. For the reliable mode it actually has assets and mandates, so it does distribute transaction commits on the nodes, and behaves in an asset way.”

The Test Parameters
In order to ensure a fair and unbiased evaluation of the databases, Thumbtack used the Yahoo! Cloud Serving Benchmark to test the databases in a write all, read one model of consistency to determine the maximum throughput for each database, either with a super balanced read write work load, or in a 95% read 5% write work load.

“For our test, we used the write all, read one model of consistency for our test, because it is durable, for sure. But, it’s something to keep in mind when you’re trying to design your database.”

“The first thing we did was break these performance based lines into two different scenarios. We called them ‘fast’ and ‘reliable.’ Fast essentially meant that we’re going to configure these databases to serve as much traffic as they can, without regard to your ability to obtain data. The data still gets written to disk, you know, it still gets replicated to other nodes. It just doesn’t do it synchronously. Reliable is simply the converse. We’re just going to take the data, make sure it gets to SSD, and not consider any writes valid until it’s written to disk, and replicated.”

“The tests are actually very simple. We put the databases on the cluster. We tuned the crap out of them, and made sure everything is optimized with the appropriate hardware, according to the vendor recommendations. All these guys are our partners. They all helped us, you know, make sure everything was tuned, and yelled at us a lot when we made mistakes. And then we the loaded the data set to the disk, just doing regular inserts, determined the maximum throughput we can get, either with a super balanced read write work load or in a 95% read, 5% write workload. And then do that again, and there’s your latency for various levels of traffic. And then repeat the whole thing for the end memory, the fast case.”

The Challenge
“This was much, much harder to do than we thought because these databases are different in terms of how you size your data set. Fundamentally, they work in very different ways. They take very different steps to achieve [their outcomes], and it’s very hard to compare them on equal terms.”

Stay tuned for the next two blog posts which will dive into the results of the benchmark and how latency affected throughput results.