TidalScale Overview
4 likes1,025 views
Tidalscale addresses the challenges of scaling applications in an era of increasing data complexity and volume by providing a solution that aggregates hardware resources under a unified operating system, reducing developer complexity and costs. Their technology allows for in-memory data management and real-time performance without the need for application modifications, making it suitable for sectors like data mining and bioinformatics. Tidalscale enables organizations to scale their hardware seamlessly, ensuring that applications can evolve alongside their needs while maintaining optimal performance.
TidalScale Overview
- 1. Scale | Simplify | Optimize | Evolve RESTORING DEVELOPER PRODUCTIVITY THROUGH SIMPLICTY 3/21/2016 COPYRIGHT 2014 TIDALSCALE, INC. 1
- 2. Industry Context Data sizes for data under management are monotonically increasing ◦ Who wants less data? Our appetite for analysis is monotonically increasing ◦ Do you think, or do you know? ◦ Trend toward evidence-based management Our appetite for speed is monotonically increasing ◦ Who wants questions answered more slowly? ◦ Hence the industry interest in in-memory data management systems Our overall ability to manage complexity is not increasing 3/21/2016 COPYRIGHT 2014 TIDALSCALE, INC. 2
- 3. Virtualization today: Is about Single Node Resource Sharing 3/21/2016 COPYRIGHT 2014 TIDALSCALE, INC. 3 Solution scaling constrained by physical machine cost and size Result: Finite resource pool
- 4. Scale-Out today: An Architecture of Developer Complexity 3/21/2016 COPYRIGHT 2014 TIDALSCALE, INC. 4 Solution scaling is constrained by the programmers ability to comprehend and navigate the architectural complexity Result: Increased development cost
- 5. TidalScale enables: An Architecture of Simplicity – One Operating System 3/21/2016 COPYRIGHT 2014 TIDALSCALE, INC. 5 Solution scaling enabled by aggregating hardware resources via one OS instance Result: Lower development complexity, lower development cost
- 6. Why? Creates a unique scalable solution experience: User experience bare metal User experience TidalScale 3/21/2016 COPYRIGHT 2014 TIDALSCALE, INC. 6 Real Screen Shots No API’s to learn – define the machine size - boot it up
- 7. Fault Tolerance System Standby and Linux Tool Transparency 3/21/2016 COPYRIGHT 2014 TIDALSCALE, INC. 7 Active Server Standby Server Transparent access to Linux Availability Tool Chain TidalScale handles the underlying complexity
- 8. Result AnyLogic™ 1 Million Agent Simulation* Normal run time - 3 Days TidalScale run time - 30 minutes We changed the customers approach to using AnyLogic™ and rapid prototyping of simulations 3/21/2016 COPYRIGHT 2014 TIDALSCALE, INC. 8 *No code changes were required
- 9. What Kinds of Problems Benefit? Data Mining / Finance ◦ High Data Volumes, Large Analytics, Risk Analysis, Fraud Detection, Graph Analytics using Alternative Data Sources, Risk modeling, High Frequency trading, Complex Event Modeling Bioinformatics • Next Generation DNA sequencing, Meta-genomic analysis, Finite Element Brain Modeling, Time-Series MRI Neuro-Imaging IT / Operational Systems ◦ In-House Applications, Web Controllers & Servers, Gateways, Image serving, Ad serving, OLTP, ERP, Business Intelligence 3/21/2016 COPYRIGHT 2014 TIDALSCALE, INC. 9
- 10. Data Mining: High Data Volumes Scenario ◦ Analytic app running on popular SQL database ◦ Well established – trained users, tied to diverse apps & tools ◦ Running well on a single system, except… Problem ◦ …Data volumes steadily growing… ◦ …Frequently upgrading hardware… ◦ …About to hit a wall The TidalScale Solution ◦ Scale up existing application & move more data into memory cache for real time performance ◦ Avoid re-architecting DB & rewriting applications for scale out/cluster ◦ Avoid expensive maintenance cost of a cluster as shape of data changes or use cases & query types evolve 3/21/2016 COPYRIGHT 2014 TIDALSCALE, INC. 10
- 11. Data Mining: Large Scale Analytics Scenario ◦ Analytic app running on popular SQL database ◦ Well established – trained users, tied to diverse apps & tools ◦ Proven, reliable application needing no changes, other than the need to increase capacity Problem ◦ Number of users and connections is growing ◦ Each user is generating large, demanding queries ◦ Frequently upgrading hardware ◦ About to hit a wall – add a second, third copy and offload users? ◦ But then you have a copy/synch problem, especially if data ever gets updated The TidalScale Solution ◦ Assemble a bigger computer to move more data into memory cache & take advantage of more CPUs ◦ Avoid re-architecting DB for scale out/cluster ◦ Avoid administering multiple servers & users, and copying/synching data 3/21/2016 COPYRIGHT 2014 TIDALSCALE, INC. 11
- 12. Data Mining: Mortgage Risk Analysis Scenario ◦ New risk assessment tools for loans, mortgages ◦ Use of “alternative data” in addition to credit bureaus Problem ◦ Mortgage application workload gets “peaky” as consumers react & play the interest rate game ◦ Backlogs develop during peaks, loan processing still remarkably hands-on process. When backlogs build, customers leave TidalScale Solution ◦ Enables a big, flexible computer that scales to handle more complex processing ◦ Supports processor intensive analytics of unstructured data, example NLP (natural language processing) ◦ Easy system expansion as data volumes grow from the addition of unstructured data 3/21/2016 COPYRIGHT 2014 TIDALSCALE, INC. 12
- 13. Data Mining: Fraud Detection Scenario ◦ New types of tools to detect fraud ◦ Ex: Graph DBs to see coordinated activity from disparate data Problem ◦ Graph databases inherently require a closely coherent (single system) view ◦ As system grows, you must either ◦ Split the graph – at the risk of losing potential connections – leaves places for bad guys to hide ◦ Reduce granularity of data – abstracting your view – loss of detail TidalScale Solution ◦ Enable full detailed view of transactional and account data across all customers over time ◦ Run the graph in-memory for real-time performance ◦ Run system on entirely standard hardware, OS & operational infrastructure 3/21/2016 COPYRIGHT 2014 TIDALSCALE, INC. 13
- 14. What enables the solution? The Memory Hierarchy in Human Terms 3/21/2016 COPYRIGHT 2014 TIDALSCALE, INC. 14 (.3ns = 1s) Event Latency Scaled 1 CPU Cycle 0.3 ns 1 s Level 1 Cache Access 0.9 ns 3 s Level 2 Cache Access 2.8 ns 9 s Level 3 Cache Access 12.9 ns 43 s Main Memory Access (DRAM, from CPU) 50.0 ns 3 min Memory over Ethernet 3.2 μs 3.2 hours CPU Context State Transfer 6.0 μs 6.0 hours Flash SSD (PCI-e) 4.7 ms 5 months Rotational disk I/O 1-10 ms 1-12 months Internet: San Francisco to New York 40 ms 4 years Internet: San Francisco to United Kingdom 81 ms 8 years Internet: San Francisco to Australia 183 ms 19 years TCP packet retransmit 1-3 s 105-317 years
- 15. TidalScale Completely Transparent - No Changes Required 3/21/2016 COPYRIGHT 2014 TIDALSCALE, INC. 15 …and many other applications
- 16. Hardware Example System features • Admin node 1 • Worker nodes 25 • Total Memory 3.2TB • Total Cores: 150 • Network 1/10GbE • Storage FreeNAS, xTB Components • 1x Admin node (A003) • Colfax CX1260i-X6 Haswell, E5-2603V3 6C, 16GB • 25x Worker nodes • Colfax CX1260i-X6 Haswell, E5-2603V3 6C, 128GB • 1x 1G switch • 2x 10G switch (S009, S010) Mellanox • 1x NAS 163/21/2016 COPYRIGHT 2014 TIDALSCALE, INC.
- 17. TidalScale Benefits Summary In-Memory Performance ◦ The world wants data to be in-memory, but hasn’t been able to get it. ◦ Historically the industry has scaled applications to fit on available computer hardware. Now, for the first time, the industry can scale the hardware to fit the application. Linear System Scalability ◦ TidalScale makes it possible to organically grow hardware using low cost commodity servers at linear cost, as customer needs evolve. ◦ Applications can now achieve superior in-memory performance using inexpensive unmodified hardware. Reduced Software Development Costs ◦ TidalScale uses off-the-shelf, unmodified Linux. ◦ TidalScale requires no changes to applications or database software. ◦ Optimization happens automatically using our software, and learns over time. 3/21/2016 COPYRIGHT 2014 TIDALSCALE, INC. 17
- 18. Price/Performance at Scale 3/21/2016 COPYRIGHT 2014 TIDALSCALE, INC. 18 Scale up costs increase rapidly after 3Tb…
- 19. Performance Scales Up 3/21/2016 COPYRIGHT 2014 TIDALSCALE, INC. 19 TidalScale can continue to add cores as well as memory (and bandwidth) Single machine solution can only add memory
- 20. Scale, Simplify, Optimize, Evolve Scale: ◦ Aggregates compute resources for large scale in-memory analysis and decision support ◦ Scales like a cluster using commodity hardware at linear cost ◦ Allow customers to grow gradually as their needs develop Simplify: ◦ Dramatically simplifies application development ◦ No need to distribute work across servers ◦ Existing applications run as a single instance, without modification, as if on a highly flexible mainframe Optimize: ◦ Automatic dynamic hierarchical resource optimization Evolve: ◦ Applicable to modern and emerging microprocessors, memories, interconnects, persistent storage & networks 3/21/2016 COPYRIGHT 2014 TIDALSCALE, INC. 20
- 21. Scale | Simplify | Optimize | Evolve RESTORING DEVELOPER PRODUCTIVITY THROUGH SIMPLICTY 3/21/2016 COPYRIGHT 2014 TIDALSCALE, INC. 21