Upload
others
View
4
Download
0
Embed Size (px)
Citation preview
Case Study SAP HANA Accelerating business with In-Memory transactions
Businesses must make smart decisions to stay ahead of the competition. One wrong decision can set a company back months. To reduce risk, businesses have turned to data analytics to examine the variables of their products, customers, partners, and even regulatory and environmental influences to make informed decisions. But the glut of data is overwhelming and the rate of change is beyond the ability of legacy processing systems to keep up. Delayed analysis makes getting current information nearly impossible: there’s so much to gather and analyze that results can be outdated before they come in.
Increasing volumes of data plus growing numbers of data sources combine with ordinary processing methods to result in high information latency. That means:
• Poor Visibility: planning based on outdated info
• Low Agility: sub-‐optimal execution of business plans
• Missed Opportunities: “delayed reaction” business model
WARP Mechanics and SAP, providers of business intelligence platforms, have developed an advanced system for business analytics that reduces the overhead of transaction processing to deliver actionable information to decision makers. SAP HANA, the high-‐performance analytic appliance, processes massive amounts of real time data in memory to provide immediate results from analysis, regardless of what data sources are involved. This system uses data compression and partitioning with real-‐time capture, insert on delta, and replication, to streamline dataflows. The WARP MemoryMatrix appliances provide the low-‐latency persistent memory storage for unprecedented performance and scalability not found in legacy storage systems, at a price point other solid-‐state providers can’t match.
In-‐memory computing, plus data-‐source independence, plus gigabytes of high IOPS throughput overcomes the delays of legacy analytics engines. This provides:
• Real-‐time Information: current information for the best planning
• Agile Performance: simulate decisions on the fly based on actual data
• Competitive Advantage: optimized focus and effective spending
Performance The HANA In-‐Memory Computing Engine software bundled with WARP Mechanics hardware includes tools for data modeling and life cycle management, security, and operations, and supports multiple interfaces to extract value from any data source and send to any data consumer. In-‐memory computing executes all transactions, transformations, and complex data processing. All applications (ERP and BW) run on data residing in-‐memory and operations work on data in real time.
But to accomplish real-‐time reliable operations, the system requires persistent storage as well as RAM. The HANA persistency layer leverages the power of the WARP MemoryMatrix appliance: a scalable all-‐SSD system that can deliver over 1M IOPS and over 100Gbps of network throughput per 4u shelf.
Business Intelligence Challenge
Overcome information latency resulting from increased volumes of data, the variety and types of data sources, and delays in calculation speed to process massive amounts of real time data, and provide immediate results from analysis and/or synchronous transactions.
Solution
WARP Mechanics “MemoryMatrix” storage appliances combined with the power of SAP HANA bring the perfect mix of performance and scalability to business intelligence.
RDMA access to pure SSD arrays allows the WARP platform to provide superior throughput and latency, while offloading the CPUs of attached systems. WARP appliances allow greater density for SAP HANA environments because they are built on WARP’s super computing class platforms.
Results
An SAP HANA cluster running on WARP Mechanics processes massive quantities of real time data at memory speeds, with an affordable price. This results in effective CapEx spending, while supporting business intelligence simulation of multiple variables in real time.
www.WARPmech.com
1288 Columbus Ave #176 San Francisco, CA 94133 888-WARP-MECH (+1.888.927.7632) [email protected]
Copyright © 2013 WARP Mechanics Ltd. All Rights Reserved
SAP HANA: Accelerating business with In-Memory transactions
The platform has two controllers, each with high-‐speed processor and DRAM. The all-‐SSD system removes the need for a separate read cache for active files, leaving more RAM for advanced features such as clustering, replication, de-‐duplication, thin provisioning, and snapshots.
SSD modules achieve maximum write and read performance while still supporting high capacity. Each flash module can sustain ~500MBps, for high throughput designs with <3ms response time latency. Each 2TB SSD exceeds 100,000 read and write IOPS per module, making the MemoryMatrix easily capable of >1 Million IOPS per shelf.
This extreme performance removes constraints on analyzing large data sets. The system is ideal for data mining and predictive analytics, from both structured and unstructured data sources.
Scale Since data can come from any source, SAP HANA needs a storage subsystem that can grow and flex as business needs change. A single MemoryMatrix can start as small as 20TB and grow to 120TB in a single shelf.
Legacy HDD-‐based storage solutions cannot match this performance density. Even few pure memory vendors can approach either the performance or capacity density, and none can approach the price:performance ratio. If business needs require additional storage, legacy OEMs will gladly wheel in another rack full of hardware. In contrast, the WARP system can scale to a petabyte of low-‐latency SSD per rack, without downtime or performance loss.
Integrity Fast response is nice. But fast and correct answers are much more useful. History demonstrates that even a single-‐digit miscalculation can bring down rockets, buildings, and bridges. Even if the data is committed to disk reliably, there is no guarantee it will stay that way: many failure cases corrupt data within legacy RAID systems, including write holes, silent data corruption, bit rot, and simultaneous disk failures.
Contemporary drive capacities have increased the occurrence of random bit-‐flipping in storage subsystems to frequent levels. Unlike other Business Intelligence systems backed by legacy OEM storage that only provide traditional RAID parity, WARP provides advanced integrity features that protect data from
corruption. It uses RAID with single-‐, double-‐, or triple-‐parity, and then goes a step further. At every phase of its lifecycle, data is protected by ECC memory, network checksums in flight, block-‐level checksums at rest, unlimited snapshots and clones, and real-‐time replication. If a bit flip occurs for any reason, it will be detected and corrected through background integrity checks, ensuring accurate data analysis and simulation both today and in the future.
Conclusion The MemoryMatrix family of appliances delivers throughput and scalability far beyond legacy OEM products for superior in-‐memory analytics with SAP HANA. Leveraging enterprise hardware and software, they supply the highest levels of throughput, scale, and data protection at a fraction of the cost. The power and performance of a pure memory appliance provides real-‐time access to process massive quantities of data at memory speeds – regardless of data source – resulting in quite simply the best preforming analytics system on the market today.