Advertisement

Aerospike TCO Vs memory-first architectures

Aerospike
Identifying business value from Big Data, IoT, M2M using the MapR big data platform
Jul. 10, 2019
Advertisement

More Related Content

Similar to Aerospike TCO Vs memory-first architectures(20)

Advertisement

Aerospike TCO Vs memory-first architectures

  1. Considering TCO
  2. 2 Proprietary & Confidential | All rights reserved. © 2019 Aerospike Inc Company • Founded: 2009, Silicon Valley, CA • Employees: ~100 employees worldwide • Customers: 250+ • Solution: Real-time database engine • Analysts: Gartner, Forrester, 451 Research Key differentiators • Unique architecture/design for Flash Drives • Large data sets with high speed interaction • Significant cost reductions • Becoming standard for real-time banking Typical real-time use cases • Instant payments • Fraud reduction • Ad placement & auctions • Recommendations • Website – enhanced user experience • Network analytics • User profile stores • RT Billing • RT Pricing • IOT (such as smart meters, self drive cars) About Aerospike
  3. 3 Proprietary & Confidential | All rights reserved. © 2019 Aerospike Inc Aerospike – engineered for performance at scale Indexes in DRAM Data on SSD • Small amount of DRAM • For the indexes only • Avoid cost and server sprawl • No cache, so no cache misses • Predictable, low latency performance on NVMe/SSD • Engineered and optimized for SSDs • SSDs around 10-15x lower cost than DRAM - $1k per TB • Much more data per node (up to 100TB) • Reads done in parallel • Writes done optimally for SSD to reduce wear-and-tear (raw I/O)
  4. 4 Proprietary & Confidential | All rights reserved. © 2019 Aerospike Inc Next Generation database An Emerging Market with Multiple Technologies Hybrid Memory Architecture Delivers Predictable Performance, Highest Availability, and Lowest TCO TCO ($) Scale TB Alternative TCO Aerospike TCO Speed TPS Scale TB Significant functional overlap - Commodity DB problem set Unique Functional Capabilities and High Value Problem Set
  5. 5 Proprietary & Confidential | All rights reserved. © 2019 Aerospike Inc Significantly Reduce TCO using Aerospike Consider storing 10TB of data which is ingested and read in real-time – single digit millisecond response • The only approach (other than using Aerospike) which could achieve this would be a memory first based solution and the data would also need to be persisted with an underlying database. Taking an in-memory based approach would look something like this. • 10TB would be replicated 3x = 30TB of unique data and likely across at least 2 data centres for HA/DR etc. = 60TB of unique data. 6x the original. • Server/Node = 500GB RAM = approx. 400GB of usable RAM per node for data storage (not persisted). • 60TB = 150 nodes (plus the underlying database tech). • Assuming $12k/server for the hardware = $5m over 3 years - then add the cost of the licenses (assuming it’s not open source). • With Aerospike it is very different. For 10TB (60TB across 2 DC’s and 3x replication), Aerospike performs the same if not better than a memory first architecture, typically with just 4 nodes per data centre = 8 nodes in total (assuming $25k/server as SSDs are used) = $900k over 3 years - a 6x difference.
  6. 6 Proprietary & Confidential | All rights reserved. © 2019 Aerospike Inc Significantly Reduce TCO using Aerospike Server/Node = 500GB RAM = approx. 400GB of usable RAM per node for data storage (not persisted). Costs are in USD$. Unique data (TBs) All data (TBs) Nodes Basic Server cost Annual maintenance Electricity, DC costs etc. FTE’s Total 1st year cost Costs over 3 years 1 6 15 180,000 36,000 36,000 150,000 402,000 846,000 2 12 30 360,000 72,000 72,000 150,000 654,000 1,242,000 3 18 45 540,000 108,000 108,000 200,000 956,000 1,788,000 4 24 60 720,000 144,000 144,000 200,000 1,208,000 2,184,000 5 30 75 900,000 180,000 180,000 300,000 1,560,000 2,880,000 10 60 150 1,800,000 360,000 360,000 400,000 2,920,000 5,160,000 20 42 300 3,600,000 720,000 720,000 500,000 5,540,000 9,420,000 Memory first / cache approach – data may also be stored using additional hardware/software Unique data (TBs) All data (TBs) Nodes Basic Server cost Annual maintenance Electricity, DC costs etc. FTE’s Total 1st year cost Costs over 3 years 1 6 6 150,000 30,000 30,000 150,000 360,000 780,000 2 12 6 150,000 30,000 30,000 150,000 360,000 780,000 3 18 6 150,000 30,000 30,000 150,000 360,000 780,000 4 24 6 150,000 30,000 30,000 150,000 360,000 780,000 5 30 6 150,000 30,000 30,000 150,000 360,000 780,000 10 60 8 200,000 40,000 40,000 150,000 430,000 890,000 20 42 10 250,000 50,000 50,000 150,000 500,000 1,000,000 Using Aerospike – all data persisted and secure
  7. 7 Proprietary & Confidential | All rights reserved. © 2019 Aerospike Inc Aerospike – at a glance • Proven and well tested database technology designed for real-time uses at scale • Sub millisecond latencies, 99th percentile • Scales to millions of read/write transactions per second • Scales to PB’s of real-time data • Low cost at scale due to use of SSD’s Vs DRAM • Mostly used in banking, telco, digital advertising, gaming companies • Trusted technology for mission critical systems • Ease of deployment, adoption, maintenance – it just works • Connectors for Spark, Kafka, Hadoop • Simple licensing model based upon unique data stored on SSDs and number of clusters
Advertisement