Memory1 For Apache Spark - v7
Igniting Apache Spark
+
MAXIMIZING SPARK PERFORMANCE AND EFFICIENCY
UP TO 256GB OF SYSTEM MEMORY PER MODULE!
MORE WORK PER SERVER
∠
∠
∠
NETWORKING INFRASTRUCTURE
MEMORY1 MAXIMIZES MEMORY PER SERVER
LINUX CONTAINERS ADD MAX PARALLELISM
EACH SPARK SERVER DOES MORE WORK
DRAM-Only Solution
VS
Save Cost
with Fewer
Servers And
Less
Networking
Infrastructure
TERABYTES OF SYSTEM MEMORY PER SERVER
SUPPORTS MANY MORE SPARK INSTANCES
Linux Container
Linux Container
Linux Container
Linux Container
Linux Container
Linux Container
Linux Container
Linux Container
Memory1 Solution
FEWER SERVERS PER CLUSTER
∠ REDUCE CAPEX AND OPEX
∠ IMPROVE SERVER EFFICIENCY
∠ ACHIEVE INFRASTRUCTURE CONSOLIDATION
Massive Memory Capacity for Apache Spark
∠∠ diablo-technologies.com
Accelerating analytics on Apache Spark…
Data is being generated at unprecedented rates and from a growing variety of sources. As a result, modern businesses are faced with a unique
opportunity and a difficult challenge. The massive streams of incoming information represent a wealth of actionable knowledge. However, the
sheer density and velocity of the incoming data, coupled with the desire for accurate, real-time analysis, makes it very challenging to process and
manage.
These requirements are driving adoption of big data processing applications like Apache Spark. Spark provides a general-purpose clustered
computing framework that rapidly ingests and processes real-time streams of data in-memory, enabling instantaneous analytics and
decision-making. With its reliance on in-memory object creation and transformation, Spark supports extremely high performance, but at the
expense of system memory usage. Due to the capacity and cost constraints of DRAM, Spark users have been forced to make undesirable design
trade-offs…until now.
…with massive memory capacity…
With Memory1, each server can support up to four times more memory than DRAM-only solutions. By intelligently leveraging high-capacity,
non-volatile memory, Memory1 modules provide huge benefits for the performance and economics of Spark clusters and enable businesses to:
DO MORE WORK PER SERVER
By dramatically increasing the system memory per server, Memory1 enables each server to support more Spark instances. By leveraging Linux
containers to encapsulate Spark environments, many Spark sessions can run on a single server in parallel. This enables each Spark node to
process significantly more data per server.
DEPLOY FEWER SERVERS PER CLUSTER
With each server able to do more work, fewer servers are required for each Spark cluster. This creates significant CAPEX and OPEX savings for new
deployments. Alternatively, in pre-existing deployments, the improved server efficiency can free additional resources for other business functions.
…and disruptive economics
The Value of Flash compared to DRAM:
Significantly Lower Cost per GB
4X Increased Module Capacity
Memory1 utilizes high-capacity NAND flash, which provides
70% Less Power per GB
significant cost advantages over DRAM. Since Memory1
DRAM Module
to traditional, DRAM-based memory solutions. With significantly larger module capacities and dramatically reduced $/GB,
Memory1 provides an unbeatable combination of business and
economic value.
Module Cost ($/GB)
modules are all-flash, they can priced very aggressively relative
8GB
16GB
32GB
64GB
128GB
256GB
Memory Module Capacity
Massive Memory Capacity for Apache Spark
∠∠ diablo-technologies.com
Was this manual useful for you? yes no
Thank you for your participation!

* Your assessment is very important for improving the work of artificial intelligence, which forms the content of this project

Download PDF

advertising