Whats the buzz about? When it comes to NoSQL, what do some of the most experienced developers know about NoSQL that makes them select Aerospike over any other NoSQL database?
Find the full webinar with audio here - http://www.aerospike.com/webinars
This presentaion will review how real-time big data driven applications are changing consumer expectations and enterprise requirements for operational databases that enable powerful and personalized customer experiences. We will describe common use cases, typical customer deployments and present an overview of Aerospike's hybrid in-memory (DRAM + Flash) and scale-out architecture.
The first round of NoSQL databases were created for point solutions to specific problems. As the space has matured, there are new, more complex use cases that is driving the evolution of NoSQL.We will go through what some of these changes are, who is using AS and why, and and architectural overview of the database.
One of the big changes in recent years has been the development of social media and multi-channel marketing. Companies that focused on a single channel have given way to companies that give their customers multiple ways to find and access content or products. These companies have focused on making it easy to share content, not only with multiple devices, but to family and friends.
Some say “Speed kills,” but in today’s world speed sells. Consumers do not have the patience to wait for a page to come up, if you can’t get them in much less than a second, you will lose them.Amazon stats - Greg Linden, Amazon https://dl.google.com/io/2009/pres/DesigningOpenSocialAppsForSpeedandScale.pdfhttp://glinden.blogspot.com/2006/11/marissa-mayer-at-web-20.htmlhttp://home.blarg.net/~glinden/StanfordDataMining.2006-11-29.pptWalmart stats - Page Performance & Site Conversion, Feb 2012“People will visit a Website less often if it is slower than a close competitor by > 250 ms” – Microsoft
Consumers now have multiple ways to access services. They do not want to have to remember on which device they did something. It should work across different devices.
Companies must be able to support customers holistically. Marketing and sales should be aware of whether an email goes out is to an existing customer, an ex-customer, or one who has been having trouble.
Customers have new expectations of when they can access systems. How often have you tried to transfer funds, only to find an issue due to “routine maintenance?”
There are many new types of offers/deals that companies can offer to customers.REAL-TIME OFFERS ONE TIME COUPONS LOYALTY REWARDSDEALS NEAR YOURELATED ITEMSRECOMMENDATIONSPRODUCT AVAILABILITYDYNAMIC PRICINGRISK SCORESWhat these all rely on is the ability to track the customer across different channels, track their behavior in real-time, track segments over a period of time (often months) and also keep track of any transactions these customers may have made. All of this requires a new infrastructure that is at once much larger and much more responsive than ever before.
To get a sense of the scale, suppose you are a medium sized business today. You might have 10 M customers you wish to track.
SaaS platform vendors (B2B2C) and consumer facing enterprises must learn from pioneers who have already built internet scale interactions platforms that deliver the right offer at the right price right now.These pioneers have all implemented the same reference architecture:An application farm backed by an Aerospike cluster, an Interaction Store that contains user information that maps identities across different browsers, devices, websites or channels, context on what the user is doing right now – clicks, swipes, searches, tweets etc and segment information from big data analytics out of a hadoop cluster or data warehouse.The hadoop clusters and data warehouses store petabytes of data while only the most recent, most valuable, 1-100TB of hot data is stored in Aerospike.Applications use identity and segment information combined with real-time context and real-time analytics to determine the right offer right now.
Companies in the ad:tech eco-system – Demand Side Platforms of DSPs, Supply Side Platforms or SSPs, Data exchanges or DMPs and Ad Exchanges – participate in Real-Time Bidding.RTB is only second to high frequency trading when it comes to low latency.These companies have just 100ms from the time someone clicks or swipes to when an offer is served.Many don’t make any money unless someone clicks on that offer – which means it has to be the right offer at the right time.They must examine the user’s cookie, check terabytes of data in Aerospike to know who the user is, what the user is doing, what the user may want, where the user is – website, mobile app, video etc – and then bid for the right to serve the offer.And only if they win the bid, can they actually serve the offer.If they cannot bid, win and serve within 100ms, they lose the opportunity and they lose revenue.
BuiltWith.com tracks which websites use various technologies.Of the top Million sites, every time hundreds of millions of users click on over 40,000 websites, a call is made to AppNexus and in turn to Aerospike.AppNexus is second only to Google and BlueKai is not far behind.This fall, AppNexus announced that Millenial Media, the largest Mobile Ad Network would be sending it’s inventory to AppNexus for real-time bidding.Microsoft just announced that it would start making ads available on outlook clients – doubling inventory and sending to AppNexus for rtb.
When it comes to speed, databases like SAP Hana go in-memory.Many databases cache data in memory and are “accelerated by flash” – but they just use SSDs instead of rotational drives.They still use the Linux file system that was built for spinning disks and get maybe a 2x performance boost.Aerospike uses a hybrid approach with indexes in DRAM and data that can be in DRAM of Flash.Aerospike can run just in DRAM alone, but most customers take advantage of the price/performance benefits of Flash.We access SSDs or PCI-E cards using a proprietary log structured file system that is optimized for flash.We use small block reads and large block writes to reduce wear and access is highly parallelized for maximum efficiency.
This results in 10x higher throughput and sub millisecond response times for these tests of Balanced reads and writesFor 2 node clusters using SSDswith 2x replicationAnd immediate consistency for Aerospike (vs eventual consistency for Cassandra)YCSB is the standard benchmark for NoSQL. Code is open sourced on github and specs are published at http://www.aerospike.com/benchmarks/ so anyone can reproduce tests and results.
High performance with Aerospike’s hybrid approach to memory also means that Aerospike can scale up to store more data per server and process the same load with 10x fewer servers than a DRAM only system.Here a customer priced a 10TB system (2x replication) that had to handle 500k TPS.A 14 server cluster with Aerospike cost only $236k compared to the 186 server cluster they would have had to purchase, install, operate and maintain with the competition.
Gartner just published it’s Magic Quadrant on Operational Databases.Established companies like Oracle, Microsoft, IBM and SAP are in the leader quadrant.Of the emerging NoSQL players,Aerospike was the ONLY Visionary. The next generation of NoSQL and the next generation of In-Memory computing.