SlideShare a Scribd company logo
1 of 31
Download to read offline
PRESENTATION TITLE GOES HERE
The Evolution, Functionality and Benefits of
NVDIMMs for Storage and Server Applications
Arthur Sainio
Bob Frey
June 30th, 2015
Agenda
Memory / Storage Hierarchy
Evolution of NVDIMMs
Taxonomy
NVM Trend/Adoption
NVM Roadmap
NVM Performance
How NVDIMMs Work
NVDIMM Features
Applications
Ecosystem and Industry Standardization
2
Memory Hierarchy
Data-Intensive applications need fast access to storage
Large performance gap between main memory and HDD
SSDs have narrowed the gap, but a ~1000X gap still exists until a
SCM becomes viable for mainstream adoption
3
Performance Gap
Adapted from SNIA presentations by Viking, HP
NVM
NAND Flash
Magnetic
DRAM
Registers
Cache
CPU
Longest Latency,
Lowest Cost
Shortest Latency,
Highest Cost
Load/Store
(Application Direct Access)
Block
(Application Indirect Access)
NVDIMM NVRAM
STT-MRAMPCM
Nanotube, Other
Acceleration
Lowest Latency
Highest Latency
ReRAM
Storage Hierarchy – Persistent Memory
4
Legacy Storage Leading Edge Storage
100 µs
1,000 µs
Tier 0
Tier 1
Tier 2
Tier 3
Caching
Analytics
Database
Mail Servers
VOD Media Streaming
Data Warehouse
Archive
Content Delivery
Backup
Surveillance
CRM
Web Hosting
OLTP
Indexing
In-memory
Hot Data
Cold Data
Access
to Host
Network
Accessible
Latency
.01 µs
10 µs
Memory Channel
NVDIMM
NVRAM
PCIe
SNIA presentations by Netlist
The Evolution of NVM
Origin: to replace battery-backed for NVM cache on storage array
controllers
NVRAM (Hybrid)
PCIe interface, stores in DRAM, backs up into NAND only on a power loss,
Supercap based, eliminates battery, maps in MMIO
NVDIMM (Hybrid)
NVDIMM-N Stores in DRAM, backs up into NAND only on a power loss,
Supercap based, eliminates battery
NVDIMM-F – maps NAND into memory address space
NVDIMM-P - maps NAND and DRAM into memory address space
And Beyond! FPGA on NVDIMM
Accelerate dynamically changing workloads
SCM – Storage Class Memories
Next Gen memory technologies (ReRAM, STT-MRAM, PCM)
Inherently persistent and removes the supercap module
5Hybrid: Memory subsystem consists of Flash and DRAM
JEDEC NVDIMM Taxonomy
NVDIMM-F
NVDIMM-N
NVDIMM-P
• One Access Method: direct byte-oriented access to DRAM and a persistence
backup/restore function on power fail
• Memory mapped DRAM. No system access to Flash
• Capacity: DRAM DIMM (8GB, 16GB, 32GB) – uses DRAM and Flash
• Latency: DRAM (10’s of nanoseconds)
• Energy source for backup
• JEDEC type and electrical mechanical definition completed
• One Access Method: block-oriented access through a shared command buffer, i.e. a
mounted drive.
• Memory mapped Flash. DRAM is not system mapped
• Capacity: NAND (100’s GB – 1’s TB) - uses NAND flash; the command buffer may be
DRAM, SRAM, etc.
• Latency: NAND (10’s of microseconds)
• JEDEC type definition completed
• Two Access Methods: persistent DRAM (–N) and also block-oriented drive access (–F)
• Memory mapped Flash and memory mapped DRAM
• Supported -> Load/Store, Emulated Block
• Capacity: NVM (100’s GB – 1’s TB) – uses DRAM and NAND flash
• Latency: NVM (100’s of nanoseconds)
• No JEDEC definition yet
Single letter designator - combines the media technology (NAND, etc) and the
access mechanism (byte, block, etc.)
Adapted from SNIA presentations by JEDEC, HP
7
NVDIMM Market Trends
• NVDIMM market expected
to reach $100’s of millions in 3-4
years
• Multiple NVDIMM vendors
• NVDIMMs are used in server and
storage applications (data centers
and cloud computing
• The market may further take off as
the Software ecosystem matures
• Industry work underway to
develop the load/store stack
• The market may follow the PCIe
trend starting 2018 with 3.3X
growth, 2018-$0.6B to 2021-$5.5B
0
50
100
150
200
250
2014 2015 2016 2017 2018 2019
200GB
400GB
800GB
0
20
40
60
80
100
120
140
160
180
200
2014 2015 2016 2017 2018 2019
8GB
16GB
NVDIMM-N Forecast (K/U)
NVDIMM-F Forecast (K/U)
Source; IHS, Netlist, Mar 2015
$600M
$5.5B 2018
8
NVM Adoption by System Application
8
DRAM NAND
+
StorageStorageStorageStorage
RAID Controller
PCIe Slot
NVDIMM
Memory
Memory
Memory
Memory
CPU
StorageStorageStorageStorage
NVM
PCIe Slot
Memory
Memory
Memory
Memory
CPU
StorageStorageStorageStorage
Memory
Memory
Memory
NVDIMM
CPU
Rack Scale Architecture
SNIA presentations by Netlist
NVvault®
NV-RDIMM (Type-N)
Non-JEDEC Standard
NVM Roadmap – Enterprise Storage –
2DPC
9
Tick
Haswell / Grantley
1.2 V 2133 MT/s
16 GB DDR4 RDIMM
Tock
Grantley
1.2 V 2400 MT/s
16 GB DDR4 RDIMM
DDR4 NVDIMM-N
1.2 V 2400 MT/s
16 GB - RDIMM
Tick
1.2 V 2667 MT/s
32 GB DDR4 R/LRDIMM
DDR4 NVDIMM-N
1.2 V 2DPC 2667 MT/s
32 GB Type N – R/LRDIMM
DDR3 / DDR4
Cost Crossover
JEDEC Standard
2H’14 1H’15 2H’15 1H’16 2H’16 1H’17
Product Availability
8/16GB-
R
32GB-L
DDR4 NVDIMM-N
1.2 V 2133 MT/s
8 GB - RDIMM
DDR4 NVDIMM SPD DDR4 Gen2 LRDIMM Raw Cards
NV Register Set
DDR4 Gen2 LRDIMM Chipset
8Gb DDR4 2400 MT/s DDR4 Gen2 LRDIMM Chipset
DDR4 Gen2 LRDIMM Raw Cards
Validated
Spec Closed Spec Closed
Spec Closed
Spec Closed
DDR4 NVDIMM-F
1.2 V 2400 MT/s
400 GB - RDIMM
DDR4 NVDIMM-P
1.2 V 2667 MT/s
16 GB Type N + 400 GB Type F
R/LRDIMM
NV-LRDIMM (Type-N, -F, -P)
JEDEC Standard
SNIA presentations by Netlist
Netlist 8 GB DDR3
2DPC 1600
Performance Benchmark –
NVDIMM- Type N DDR3 - IOPs & Response
10
Performance - IOPs
Symmetric Read & Write Performance (1.6M IOPS)
Very low Average Response Times (0.02 mS)
Bandwidth – MB/s
Very Fast Throughput (70GB/s Read / 21GB/s Write- Random)
Very Low Average Response Time (1.8 mS R, 5.8 mS W)
Response Time (Latency at single OIO or T1Q1)
RND 4K 100% Write - .004 mS Average Response Time
RND 4K 100% Write - .120 mS Maximum Response Time
SNIA presentations by Netlist
Form factor is
RDIMM/LRDIMM JEDEC
module
Device model is “SSD on
DDR bus”; that is, storage
device not a memory device
Typically visible to system
as block storage device
Flash capacity exposed
directly to host, as in any
SSD 11
NVDIMM(-F) – How it Works
NVDIMM-F
Controller
Flash
Host Driver
MCU
Host
Device
DDRBus
Adapted from SNIA presentations by SanDisk
RDIMM/LRDIMM JEDEC module with power backup
Data stored into NAND on power loss to
achieve persistence
Host only addresses the DRAM and has
no direct access to the flash
NVDIMM controller moves data from
DRAM to flash upon power loss or other
trigger; can back up portions or all of
DRAM upon command
When power fails a backup power source
provides power to the NVDIMM while DRAM
is backed up to Flash
MRC (Memory Reference Code) configures NVDIMM
controller to move data back from Flash to DRAM when
recovery is needed
12
Backup Power
NVDIMM
MCU
NVDIMM(-N) – How it Works
Form factor is RDIMM/LRDIMM
JEDEC module
Three Modes of Operation
Storage Mode: “SSD on DDR bus”;
Storage device not a memory device,
No application SW changes needed,
BIOS & Driver required.
Memory Mode: DRAM Operation,
Volatile
NVM Mode
Hybrid Mode: Combines NVM and
Storage.
13
NVDIMM(-P) – How it Works
NVDIMM-
Controller
Flash
Host Driver
MCU
Host
Device
DDRBus
DRAM
NVDIMM (–N) Features
NV (Digital) Controller
• Facilitates backup/restore functionality
• Register access for temp monitoring,
ultracapacitor, and flash statistics
Standard DDR4
RDIMM Interface
14
SNIA presentations by Netlist
DDR3 DDR4
Density Up to 8GB Up to16GB
Power (V) 1.5 / 1.35 1.2
Performance 1DPC 1600 MT/s
2DPC 1600 MT/s
1DPC 2400 MT/s
2DPC 2133 MT/s
Ultracapacitor Backup
• Provides power to move vital data contents
from DRAM to flash during a system or power
failure event
• Eliminates Battery
• Charged by NVDIMM via 12V power
NVDIMM –N Features
DDR3
4GB/8GB Capacity
DDR3 1600
DDR3 latency
DDR3 Dual voltage 1.35V/1.5V operation
Configuration: 1Rx8
RDIMM Interface
12V Aux via a proprietary connection
Availability - Now
DDR4
4GB/8GB/16GB Capacity
DDR4 2400
DDR4 latency
DDR4 voltage 1.2V
Configuration: 1R/2R
RDIMM (2015) & LRDIMM (2016) Interface
12V support via the DDR4 DIMM connector
Availability - Now
ALL
JEDEC Form Factor
Support in Intel MRC
Follows RDIMM Population Rules
SLC or MLC NAND
End-to-end Data integrity
Backup initiated via SAVE Signal or I2C
OS SAVE as part of shutdown
SAVE during operation requires system
re-initialization of DDR3 register & re-write
DDR3 mode register
Backup duration
4GB ~ 16GB (Vendor Specific)
Health Monitoring
Error reporting
Predictive failure warnings
Backup, restore, power, NAND life status
JEDEC or Proprietary
15
SNIA presentations by Netlist
NVDIMM Stack
Adapted from SNIA presentations by AgigA Tech
Platform Hardware
NVDIMM
BIOS
OS
CPU/Memory Controller
NVDIMM
Power Supply
SMB
MRC + BIOS Modules
DIMM Interface
Energy Module
SAVE Trigger
NVDIMM Driver
Block Driver
Application
Byte Addressable
Load StoreBlock Mode
Software
Hardware
Kernel Space
User Space
21
Application Example: HP NVDIMM-N
(DDR3) Acceleration – by HP
17
4X the Acceleration inThroughput and Latency
Source: SNIA 2/15
Adapted from SNIA presentations by HP
Application Example:
Storage Bridge Bay (SBB)
CPU
DDR3
NVM
PCIe
Card
10GigE
NIC
DDR3
DDR3
DDR3
CPU
NVM
PCIe
Card
10GigE
NIC
DDR3
DDR3
DDR3
DDR3
Backplane
PCIe
PCIe
PCIe
PCIe
Shadow Writes Required for Failover
18
Adapted from SNIA presentations by AgigA Tech
SBB: A Simpler/Better/Faster Way
CPU
DDR3
DDR3
DDR3
NVDIMM
DDR3
DDR3
DDR3
NVIDIMM
Backplane
Non-Transparent Bridge
(PCIe)
CPU
19
Also a better alternative to Cache-to-Flash implementations:
• Separate failure domain
• No battery maintenance
• System hold-up requirements significantly less severe
• 4x write latency performance improvement
Adapted from SNIA presentations by AgigA Tech
Tier 0
25ns
NVDIMM
Type 1 (N,F,P)
Tier 1
100us PCIe SSD
15us PCIe NVM
Tier 2
100us
SATA/SAS
NVDIMM
NVRAM
Store Metadata in memory for application acceleration
Host Caching: As Cache for direct attached PCIe SSD
Fast RAID Computation as block device for distributed storage
Check-pointing state for fast sync and restore
SSD Mapping Table
Persistent RAM Disk
As Fast 4K block store
Store boot image for fast restore
Application Uses – Storage Main
Memory & PCIe
20
PersistentVariables
Metadata
Checkpoint State
Host Caching
RAMDisk
RAID Compute
Write buffer
SSD Mapping
Journaling
Logging
SNIA presentations by Netlist
Advantages of NVDIMMs for
Applications
Legacy HDD/SSD Solution
Persistent data stored in HDD or SSD tiers
Slow & unpredictable software stack
NVDIMM Solution
Persistent data stored in fast DRAM tier
Removes software stack from data-path
Accelerates SW-Apps !
• DRAM class latency & thru-put for persistent data
– 1000X lower latency
– 10X+ throughput increase
– But, 10X lower capacity vs. SSD
• The value is in application acceleration
Kernel Module
Simplification
RAID & Storage
Tiering
Write Buffers
Persistent
Caches
Kernel
Optimization
Transaction
Logs

21
Low Write-Latency Persistent Storage
Value Proposition Specifics
• Accelerating Write-Latency
Datacenter Workloads
• Transaction Commits, Logs & Journals
400GB SLC PCIe SSD 8GB NVDIMM Advantage
Write Latency (us) >15 <0.1 >150x
Read Latency (us) >47 <0.1 >470x
Endurance Workload-dependent Unlimited Unlimited
PCIe Lanes Consumed 8 0 No PCIe used
Power (W) <25 <10 >2x
 >100x Lower Write Latency
 Unlimited Endurance
 Scalable
 No PCIe Resources Consumed
 Lower Total Cost of Ownership
• $/Latency, $/IOPS & IOPS/W
Adapted from SNIA presentations by Inphi
28
$0M
$500M
$1,000M
$1,500M
$2,000M
$2,500M
$3,000M
$3,500M
$4,000M
CAPEX OPEX TOTAL
SATA SSD x2 NV3 x1
$0M
$200M
$400M
$600M
$800M
$1,000M
$1,200M
$1,400M
$1,600M
$1,800M
$2,000M
CAPEX OPEX TOTAL
PCIe SSD x1 NV3 x1
User Case - 5yr TCO Analysis
SATA/PCIe vs. NVDIMM
23
100,000 Servers 50,000 Servers 1,000 Servers
98% reduction
$2,723M saved
98% reduction
$695M saved
98% reduction
$3,418M saved
96% reduction
$1,771M saved
96% reduction
$1,424M saved
96% reduction
$347M saved
Dell PowerEdge E-26xx CPU
DDR3 NVDIMM 1600 MTs
Write Intensive App
98%
reduction
96%
reduction
NVDIMM(-N) Ecosystem
NVDIMMs &
Systems
New
Applications
Open Source
Drivers & User
API
BIOS, MRC
Platform
Software
Mass Deployment
Hardware
Standardization
• System management,
Power health
• System support H/W
trigger (ADR)
• Mechanical (power
source)
• JEDEC NVDIMM
Platform Support
• Off-the-shelf and
OEM platform
support for NVDIMM
today
• System supported
H/W trigger (ADR)
• Mechanical (power
source)
Software
Standardization
• Applications
• Linux NVDIMM-
aware kernel 4.1
• API’s
BIOS Support
• NVDIMM-aware
BIOS
• Intel modifications to
MRC to support
NVDIMMs
• JEDEC NVDIMM
I2C command set
• JEDEC SPD 30
NVDIMM(-N) Standardization
JEDEC Hybrid Memory Task Group
DDR4 12V Power Pins (1, 145) standardized
DDR4 SAVE_n Pin (230) standardized
12V in DDR4 socket will simplify NVDIMM power circuitry and cable
routing
Under discussion: Standard system interface for NVDIMM
i2c register map for NVDIMM-N controller
Issued ballot on NVDIMM Controller Event Pin
SPD for NVDIMM representation
SNIA NVDIMM SIG
(Special Interest Group, >20 companies)
Formed in 2014 as a SIG of the SNIA Solid State Storage Initiative
Communicating existing industry standards, and areas for vendor
differentiation
Helping technology and solution vendors whose products integrate
NVDIMMs to communicate their benefits and value to the greater
market
Developing vendor-agnostic user perspective case studies, best
practices, and vertical industry requirements
Standards
JEDEC
Flow of NVDIMM Adoption and Support
BIOS
NVDIMM
Suppliers
Motherboard
ODMs. OEMs
Platform
Integrated
Solutions
31
Eliminate File System Latency
with Memory Mapped Files
Application
File System
Disk Driver
Disk
Application
Persistent
Memory
Load/Store
Memory Mapped Files
Traditional New
UserKernelHW
UserHW
Courtesy: 2015 Data Storage Innovation Conference. © 2015 Storage Networking Industry Association. All Rights Reserved.
Conventional Block and File Modes
Application
NVM block capable driver
File system
Application
NVM device NVM device
User space
Kernel space
Native file
API
NVM.BLOCK mode
NVM.FILE modeUse with disk-like NVM
NVM.BLOCK Mode
 Targeted for file systems and block-
aware applications
 Atomic writes
 Length and alignment granularities
 Thin provisioning management
NVM.FILE Mode
 Targeted for file based apps.
 Discovery and use of atomic write
features
 Discovery of granularities
Courtesy: 2015 Data Storage Innovation Conference. © 2015 Storage Networking Industry Association. All Rights Reserved.
Persistent Memory Modes
Application
PM device PM device PM device. . .
User space
Kernel space
MMU
MappingsPM-aware file system
NVM PM capable driver
Load/
store
Native file
API
PM-aware kernel module
PM device
NVM.PM.VOLUME mode
NVM.PM.FILE mode
Use with memory-like NVM
NVM.PM.VOLUME Mode
 Software abstraction to OS
components for Persistent Memory
(PM) hardware
 List of physical address ranges for
each PM volume
 Thin provisioning management
NVM.PM.FILE Mode
 Describes the behavior for
applications accessing persistent
memory Discovery and use of atomic
write features
 Mapping PM files (or subsets of files)
to virtual memory addresses
 Syncing portions of PM files to the
persistence domain
Courtesy: 2015 Data Storage Innovation Conference. © 2015 Storage Networking Industry Association. All Rights Reserved.
Building on the Basic PM Model
NVM.PM.FILE
programming model
“surfaces” PM to
application
Refine API with additional
libraries that evolve into
language extensions
Add compatible
functionality to PM file
systems
Courtesy: 2015 Data Storage Innovation Conference. © 2015 Storage Networking Industry Association. All Rights Reserved.
Open Source Sub-initiatives
30
Initiative What will it do? What is the status?
1 NVDIMM Aware Linux Kernel
Support for NVDIMM-N modules
(arch/x86/kernel/e820.c)
Motherboard vendors provide BIOS/MRC
changes needed to recognize NVDIMM-N
modules.
Customers still need either Block or Load/Store
Linux driver to enable NVDIMM-N modules.
Waiting for the ACPI Spec 6.0 to be
published.
Availability in Github - April/2015.
2 Block & Load/Store Drivers
Vendor agnostic block mode and load
store drivers for NVDIMM-N modules.
In addition, provides support for DAX
(direct access) functionality.
The load/store driver will likely come after the
block driver into the Linux Kernel.
The PMEM initiative is awaiting ACPI
related changes to be approved.
General Availability in Linux Kernel - TBD
3 DAX Support
Direct Access (DAX) support for
NVDIMM-N modules in Ext4
Git: https://github.com/01org/prd
Path: fs/ext4/* fs/block_dev.c
The Ext4 file system DAX support to NVDIMM-N
modules eliminates the page cache layer
completely.
This requires the availability of [2] & [3].
Q2, 2015, in the official 3.20 kernel
PRESENTATION TITLE GOES HERE
Thank You!
35

More Related Content

Viewers also liked

IMC Summit 2016 Keynote - Arthur Sainio - NVDIMM: Changes are Here So What’s ...
IMC Summit 2016 Keynote - Arthur Sainio - NVDIMM: Changes are Here So What’s ...IMC Summit 2016 Keynote - Arthur Sainio - NVDIMM: Changes are Here So What’s ...
IMC Summit 2016 Keynote - Arthur Sainio - NVDIMM: Changes are Here So What’s ...In-Memory Computing Summit
 
Hands-on Lab: How to Unleash Your Storage Performance by Using NVM Express™ B...
Hands-on Lab: How to Unleash Your Storage Performance by Using NVM Express™ B...Hands-on Lab: How to Unleash Your Storage Performance by Using NVM Express™ B...
Hands-on Lab: How to Unleash Your Storage Performance by Using NVM Express™ B...Odinot Stanislas
 
Intel, Micron unveil “breakthrough” 3D XPoint Memory Tech – A revolutionary b...
Intel, Micron unveil “breakthrough” 3D XPoint Memory Tech – A revolutionary b...Intel, Micron unveil “breakthrough” 3D XPoint Memory Tech – A revolutionary b...
Intel, Micron unveil “breakthrough” 3D XPoint Memory Tech – A revolutionary b...Syntech
 
Operating Systems 1 (9/12) - Memory Management Concepts
Operating Systems 1 (9/12) - Memory Management ConceptsOperating Systems 1 (9/12) - Memory Management Concepts
Operating Systems 1 (9/12) - Memory Management ConceptsPeter Tröger
 
Intel and DataStax: 3D XPoint and NVME Technology Cassandra Storage Comparison
Intel and DataStax: 3D XPoint and NVME Technology Cassandra Storage ComparisonIntel and DataStax: 3D XPoint and NVME Technology Cassandra Storage Comparison
Intel and DataStax: 3D XPoint and NVME Technology Cassandra Storage ComparisonDataStax Academy
 
Moving to PCI Express based SSD with NVM Express
Moving to PCI Express based SSD with NVM ExpressMoving to PCI Express based SSD with NVM Express
Moving to PCI Express based SSD with NVM ExpressOdinot Stanislas
 
CTI Group- Blue power technology storwize technical training for customer - p...
CTI Group- Blue power technology storwize technical training for customer - p...CTI Group- Blue power technology storwize technical training for customer - p...
CTI Group- Blue power technology storwize technical training for customer - p...Tri Susilo
 

Viewers also liked (8)

IMC Summit 2016 Keynote - Arthur Sainio - NVDIMM: Changes are Here So What’s ...
IMC Summit 2016 Keynote - Arthur Sainio - NVDIMM: Changes are Here So What’s ...IMC Summit 2016 Keynote - Arthur Sainio - NVDIMM: Changes are Here So What’s ...
IMC Summit 2016 Keynote - Arthur Sainio - NVDIMM: Changes are Here So What’s ...
 
Hands-on Lab: How to Unleash Your Storage Performance by Using NVM Express™ B...
Hands-on Lab: How to Unleash Your Storage Performance by Using NVM Express™ B...Hands-on Lab: How to Unleash Your Storage Performance by Using NVM Express™ B...
Hands-on Lab: How to Unleash Your Storage Performance by Using NVM Express™ B...
 
Intel, Micron unveil “breakthrough” 3D XPoint Memory Tech – A revolutionary b...
Intel, Micron unveil “breakthrough” 3D XPoint Memory Tech – A revolutionary b...Intel, Micron unveil “breakthrough” 3D XPoint Memory Tech – A revolutionary b...
Intel, Micron unveil “breakthrough” 3D XPoint Memory Tech – A revolutionary b...
 
Operating Systems 1 (9/12) - Memory Management Concepts
Operating Systems 1 (9/12) - Memory Management ConceptsOperating Systems 1 (9/12) - Memory Management Concepts
Operating Systems 1 (9/12) - Memory Management Concepts
 
Persistent memory
Persistent memoryPersistent memory
Persistent memory
 
Intel and DataStax: 3D XPoint and NVME Technology Cassandra Storage Comparison
Intel and DataStax: 3D XPoint and NVME Technology Cassandra Storage ComparisonIntel and DataStax: 3D XPoint and NVME Technology Cassandra Storage Comparison
Intel and DataStax: 3D XPoint and NVME Technology Cassandra Storage Comparison
 
Moving to PCI Express based SSD with NVM Express
Moving to PCI Express based SSD with NVM ExpressMoving to PCI Express based SSD with NVM Express
Moving to PCI Express based SSD with NVM Express
 
CTI Group- Blue power technology storwize technical training for customer - p...
CTI Group- Blue power technology storwize technical training for customer - p...CTI Group- Blue power technology storwize technical training for customer - p...
CTI Group- Blue power technology storwize technical training for customer - p...
 

More from In-Memory Computing Summit

IMC Summit 2016 Breakout - Per Minoborg - Work with Multiple Hot Terabytes in...
IMC Summit 2016 Breakout - Per Minoborg - Work with Multiple Hot Terabytes in...IMC Summit 2016 Breakout - Per Minoborg - Work with Multiple Hot Terabytes in...
IMC Summit 2016 Breakout - Per Minoborg - Work with Multiple Hot Terabytes in...In-Memory Computing Summit
 
IMC Summit 2016 Breakout - Henning Andersen - Using Lock-free and Wait-free I...
IMC Summit 2016 Breakout - Henning Andersen - Using Lock-free and Wait-free I...IMC Summit 2016 Breakout - Henning Andersen - Using Lock-free and Wait-free I...
IMC Summit 2016 Breakout - Henning Andersen - Using Lock-free and Wait-free I...In-Memory Computing Summit
 
IMC Summit 2016 Breakout - Roman Shtykh - Apache Ignite as a Data Processing Hub
IMC Summit 2016 Breakout - Roman Shtykh - Apache Ignite as a Data Processing HubIMC Summit 2016 Breakout - Roman Shtykh - Apache Ignite as a Data Processing Hub
IMC Summit 2016 Breakout - Roman Shtykh - Apache Ignite as a Data Processing HubIn-Memory Computing Summit
 
IMC Summit 2016 Breakout - Nikita Shamgunov - Propelling IoT Innovation with ...
IMC Summit 2016 Breakout - Nikita Shamgunov - Propelling IoT Innovation with ...IMC Summit 2016 Breakout - Nikita Shamgunov - Propelling IoT Innovation with ...
IMC Summit 2016 Breakout - Nikita Shamgunov - Propelling IoT Innovation with ...In-Memory Computing Summit
 
IMC Summit 2016 Breakout - Matt Coventon - Test Driving Streaming and CEP on ...
IMC Summit 2016 Breakout - Matt Coventon - Test Driving Streaming and CEP on ...IMC Summit 2016 Breakout - Matt Coventon - Test Driving Streaming and CEP on ...
IMC Summit 2016 Breakout - Matt Coventon - Test Driving Streaming and CEP on ...In-Memory Computing Summit
 
IMC Summit 2016 Innovation - Derek Nelson - PipelineDB: The Streaming-SQL Dat...
IMC Summit 2016 Innovation - Derek Nelson - PipelineDB: The Streaming-SQL Dat...IMC Summit 2016 Innovation - Derek Nelson - PipelineDB: The Streaming-SQL Dat...
IMC Summit 2016 Innovation - Derek Nelson - PipelineDB: The Streaming-SQL Dat...In-Memory Computing Summit
 
IMC Summit 2016 Innovation - Dennis Duckworth - Lambda-B-Gone: The In-memory ...
IMC Summit 2016 Innovation - Dennis Duckworth - Lambda-B-Gone: The In-memory ...IMC Summit 2016 Innovation - Dennis Duckworth - Lambda-B-Gone: The In-memory ...
IMC Summit 2016 Innovation - Dennis Duckworth - Lambda-B-Gone: The In-memory ...In-Memory Computing Summit
 
IMC Summit 2016 Innovation - Steve Wilkes - Tap Into Your Enterprise – Why Da...
IMC Summit 2016 Innovation - Steve Wilkes - Tap Into Your Enterprise – Why Da...IMC Summit 2016 Innovation - Steve Wilkes - Tap Into Your Enterprise – Why Da...
IMC Summit 2016 Innovation - Steve Wilkes - Tap Into Your Enterprise – Why Da...In-Memory Computing Summit
 
IMC Summit 2016 Innovation - Girish Mutreja - Unveiling the X Platform
IMC Summit 2016 Innovation - Girish Mutreja - Unveiling the X PlatformIMC Summit 2016 Innovation - Girish Mutreja - Unveiling the X Platform
IMC Summit 2016 Innovation - Girish Mutreja - Unveiling the X PlatformIn-Memory Computing Summit
 
IMC Summit 2016 Breakout - Ken Gibson - The In-Place Working Storage Tier
IMC Summit 2016 Breakout - Ken Gibson - The In-Place Working Storage TierIMC Summit 2016 Breakout - Ken Gibson - The In-Place Working Storage Tier
IMC Summit 2016 Breakout - Ken Gibson - The In-Place Working Storage TierIn-Memory Computing Summit
 
IMC Summit 2016 Breakout - Brian Bulkowski - NVMe, Storage Class Memory and O...
IMC Summit 2016 Breakout - Brian Bulkowski - NVMe, Storage Class Memory and O...IMC Summit 2016 Breakout - Brian Bulkowski - NVMe, Storage Class Memory and O...
IMC Summit 2016 Breakout - Brian Bulkowski - NVMe, Storage Class Memory and O...In-Memory Computing Summit
 
IMC Summit 2016 Breakout - Yanping Wang - Non-volatile Generic Object Program...
IMC Summit 2016 Breakout - Yanping Wang - Non-volatile Generic Object Program...IMC Summit 2016 Breakout - Yanping Wang - Non-volatile Generic Object Program...
IMC Summit 2016 Breakout - Yanping Wang - Non-volatile Generic Object Program...In-Memory Computing Summit
 
IMC Summit 2016 Breakout - Andy Pavlo - What Non-Volatile Memory Means for th...
IMC Summit 2016 Breakout - Andy Pavlo - What Non-Volatile Memory Means for th...IMC Summit 2016 Breakout - Andy Pavlo - What Non-Volatile Memory Means for th...
IMC Summit 2016 Breakout - Andy Pavlo - What Non-Volatile Memory Means for th...In-Memory Computing Summit
 
IMC Summit 2016 Breakout - Gordon Patrick - Developments in Persistent Memory
IMC Summit 2016 Breakout - Gordon Patrick - Developments in Persistent MemoryIMC Summit 2016 Breakout - Gordon Patrick - Developments in Persistent Memory
IMC Summit 2016 Breakout - Gordon Patrick - Developments in Persistent MemoryIn-Memory Computing Summit
 
IMC Summit 2016 Breakout - Girish Kathalagiri - Decision Making with MLLIB, S...
IMC Summit 2016 Breakout - Girish Kathalagiri - Decision Making with MLLIB, S...IMC Summit 2016 Breakout - Girish Kathalagiri - Decision Making with MLLIB, S...
IMC Summit 2016 Breakout - Girish Kathalagiri - Decision Making with MLLIB, S...In-Memory Computing Summit
 
IMC Summit 2016 Breakout - Steve Wikes - Making IMC Enterprise Grade
IMC Summit 2016 Breakout - Steve Wikes - Making IMC Enterprise GradeIMC Summit 2016 Breakout - Steve Wikes - Making IMC Enterprise Grade
IMC Summit 2016 Breakout - Steve Wikes - Making IMC Enterprise GradeIn-Memory Computing Summit
 
IMC Summit 2016 Breakout - Noah Arliss - The Truth: How to Test Your Distribu...
IMC Summit 2016 Breakout - Noah Arliss - The Truth: How to Test Your Distribu...IMC Summit 2016 Breakout - Noah Arliss - The Truth: How to Test Your Distribu...
IMC Summit 2016 Breakout - Noah Arliss - The Truth: How to Test Your Distribu...In-Memory Computing Summit
 
IMC Summit 2016 Breakout - Aleksandar Seovic - The Illusion of Statelessness
IMC Summit 2016 Breakout - Aleksandar Seovic - The Illusion of StatelessnessIMC Summit 2016 Breakout - Aleksandar Seovic - The Illusion of Statelessness
IMC Summit 2016 Breakout - Aleksandar Seovic - The Illusion of StatelessnessIn-Memory Computing Summit
 
IMC Summit 2016 Breakout - Girish Mutreja - Extreme Transaction Processing in...
IMC Summit 2016 Breakout - Girish Mutreja - Extreme Transaction Processing in...IMC Summit 2016 Breakout - Girish Mutreja - Extreme Transaction Processing in...
IMC Summit 2016 Breakout - Girish Mutreja - Extreme Transaction Processing in...In-Memory Computing Summit
 
IMC Summit 2016 Breakout - Greg Luck - How to Speed Up Your Application Using...
IMC Summit 2016 Breakout - Greg Luck - How to Speed Up Your Application Using...IMC Summit 2016 Breakout - Greg Luck - How to Speed Up Your Application Using...
IMC Summit 2016 Breakout - Greg Luck - How to Speed Up Your Application Using...In-Memory Computing Summit
 

More from In-Memory Computing Summit (20)

IMC Summit 2016 Breakout - Per Minoborg - Work with Multiple Hot Terabytes in...
IMC Summit 2016 Breakout - Per Minoborg - Work with Multiple Hot Terabytes in...IMC Summit 2016 Breakout - Per Minoborg - Work with Multiple Hot Terabytes in...
IMC Summit 2016 Breakout - Per Minoborg - Work with Multiple Hot Terabytes in...
 
IMC Summit 2016 Breakout - Henning Andersen - Using Lock-free and Wait-free I...
IMC Summit 2016 Breakout - Henning Andersen - Using Lock-free and Wait-free I...IMC Summit 2016 Breakout - Henning Andersen - Using Lock-free and Wait-free I...
IMC Summit 2016 Breakout - Henning Andersen - Using Lock-free and Wait-free I...
 
IMC Summit 2016 Breakout - Roman Shtykh - Apache Ignite as a Data Processing Hub
IMC Summit 2016 Breakout - Roman Shtykh - Apache Ignite as a Data Processing HubIMC Summit 2016 Breakout - Roman Shtykh - Apache Ignite as a Data Processing Hub
IMC Summit 2016 Breakout - Roman Shtykh - Apache Ignite as a Data Processing Hub
 
IMC Summit 2016 Breakout - Nikita Shamgunov - Propelling IoT Innovation with ...
IMC Summit 2016 Breakout - Nikita Shamgunov - Propelling IoT Innovation with ...IMC Summit 2016 Breakout - Nikita Shamgunov - Propelling IoT Innovation with ...
IMC Summit 2016 Breakout - Nikita Shamgunov - Propelling IoT Innovation with ...
 
IMC Summit 2016 Breakout - Matt Coventon - Test Driving Streaming and CEP on ...
IMC Summit 2016 Breakout - Matt Coventon - Test Driving Streaming and CEP on ...IMC Summit 2016 Breakout - Matt Coventon - Test Driving Streaming and CEP on ...
IMC Summit 2016 Breakout - Matt Coventon - Test Driving Streaming and CEP on ...
 
IMC Summit 2016 Innovation - Derek Nelson - PipelineDB: The Streaming-SQL Dat...
IMC Summit 2016 Innovation - Derek Nelson - PipelineDB: The Streaming-SQL Dat...IMC Summit 2016 Innovation - Derek Nelson - PipelineDB: The Streaming-SQL Dat...
IMC Summit 2016 Innovation - Derek Nelson - PipelineDB: The Streaming-SQL Dat...
 
IMC Summit 2016 Innovation - Dennis Duckworth - Lambda-B-Gone: The In-memory ...
IMC Summit 2016 Innovation - Dennis Duckworth - Lambda-B-Gone: The In-memory ...IMC Summit 2016 Innovation - Dennis Duckworth - Lambda-B-Gone: The In-memory ...
IMC Summit 2016 Innovation - Dennis Duckworth - Lambda-B-Gone: The In-memory ...
 
IMC Summit 2016 Innovation - Steve Wilkes - Tap Into Your Enterprise – Why Da...
IMC Summit 2016 Innovation - Steve Wilkes - Tap Into Your Enterprise – Why Da...IMC Summit 2016 Innovation - Steve Wilkes - Tap Into Your Enterprise – Why Da...
IMC Summit 2016 Innovation - Steve Wilkes - Tap Into Your Enterprise – Why Da...
 
IMC Summit 2016 Innovation - Girish Mutreja - Unveiling the X Platform
IMC Summit 2016 Innovation - Girish Mutreja - Unveiling the X PlatformIMC Summit 2016 Innovation - Girish Mutreja - Unveiling the X Platform
IMC Summit 2016 Innovation - Girish Mutreja - Unveiling the X Platform
 
IMC Summit 2016 Breakout - Ken Gibson - The In-Place Working Storage Tier
IMC Summit 2016 Breakout - Ken Gibson - The In-Place Working Storage TierIMC Summit 2016 Breakout - Ken Gibson - The In-Place Working Storage Tier
IMC Summit 2016 Breakout - Ken Gibson - The In-Place Working Storage Tier
 
IMC Summit 2016 Breakout - Brian Bulkowski - NVMe, Storage Class Memory and O...
IMC Summit 2016 Breakout - Brian Bulkowski - NVMe, Storage Class Memory and O...IMC Summit 2016 Breakout - Brian Bulkowski - NVMe, Storage Class Memory and O...
IMC Summit 2016 Breakout - Brian Bulkowski - NVMe, Storage Class Memory and O...
 
IMC Summit 2016 Breakout - Yanping Wang - Non-volatile Generic Object Program...
IMC Summit 2016 Breakout - Yanping Wang - Non-volatile Generic Object Program...IMC Summit 2016 Breakout - Yanping Wang - Non-volatile Generic Object Program...
IMC Summit 2016 Breakout - Yanping Wang - Non-volatile Generic Object Program...
 
IMC Summit 2016 Breakout - Andy Pavlo - What Non-Volatile Memory Means for th...
IMC Summit 2016 Breakout - Andy Pavlo - What Non-Volatile Memory Means for th...IMC Summit 2016 Breakout - Andy Pavlo - What Non-Volatile Memory Means for th...
IMC Summit 2016 Breakout - Andy Pavlo - What Non-Volatile Memory Means for th...
 
IMC Summit 2016 Breakout - Gordon Patrick - Developments in Persistent Memory
IMC Summit 2016 Breakout - Gordon Patrick - Developments in Persistent MemoryIMC Summit 2016 Breakout - Gordon Patrick - Developments in Persistent Memory
IMC Summit 2016 Breakout - Gordon Patrick - Developments in Persistent Memory
 
IMC Summit 2016 Breakout - Girish Kathalagiri - Decision Making with MLLIB, S...
IMC Summit 2016 Breakout - Girish Kathalagiri - Decision Making with MLLIB, S...IMC Summit 2016 Breakout - Girish Kathalagiri - Decision Making with MLLIB, S...
IMC Summit 2016 Breakout - Girish Kathalagiri - Decision Making with MLLIB, S...
 
IMC Summit 2016 Breakout - Steve Wikes - Making IMC Enterprise Grade
IMC Summit 2016 Breakout - Steve Wikes - Making IMC Enterprise GradeIMC Summit 2016 Breakout - Steve Wikes - Making IMC Enterprise Grade
IMC Summit 2016 Breakout - Steve Wikes - Making IMC Enterprise Grade
 
IMC Summit 2016 Breakout - Noah Arliss - The Truth: How to Test Your Distribu...
IMC Summit 2016 Breakout - Noah Arliss - The Truth: How to Test Your Distribu...IMC Summit 2016 Breakout - Noah Arliss - The Truth: How to Test Your Distribu...
IMC Summit 2016 Breakout - Noah Arliss - The Truth: How to Test Your Distribu...
 
IMC Summit 2016 Breakout - Aleksandar Seovic - The Illusion of Statelessness
IMC Summit 2016 Breakout - Aleksandar Seovic - The Illusion of StatelessnessIMC Summit 2016 Breakout - Aleksandar Seovic - The Illusion of Statelessness
IMC Summit 2016 Breakout - Aleksandar Seovic - The Illusion of Statelessness
 
IMC Summit 2016 Breakout - Girish Mutreja - Extreme Transaction Processing in...
IMC Summit 2016 Breakout - Girish Mutreja - Extreme Transaction Processing in...IMC Summit 2016 Breakout - Girish Mutreja - Extreme Transaction Processing in...
IMC Summit 2016 Breakout - Girish Mutreja - Extreme Transaction Processing in...
 
IMC Summit 2016 Breakout - Greg Luck - How to Speed Up Your Application Using...
IMC Summit 2016 Breakout - Greg Luck - How to Speed Up Your Application Using...IMC Summit 2016 Breakout - Greg Luck - How to Speed Up Your Application Using...
IMC Summit 2016 Breakout - Greg Luck - How to Speed Up Your Application Using...
 

Recently uploaded

The Kubernetes Gateway API and its role in Cloud Native API Management
The Kubernetes Gateway API and its role in Cloud Native API ManagementThe Kubernetes Gateway API and its role in Cloud Native API Management
The Kubernetes Gateway API and its role in Cloud Native API ManagementNuwan Dias
 
UiPath Studio Web workshop series - Day 6
UiPath Studio Web workshop series - Day 6UiPath Studio Web workshop series - Day 6
UiPath Studio Web workshop series - Day 6DianaGray10
 
Igniting Next Level Productivity with AI-Infused Data Integration Workflows
Igniting Next Level Productivity with AI-Infused Data Integration WorkflowsIgniting Next Level Productivity with AI-Infused Data Integration Workflows
Igniting Next Level Productivity with AI-Infused Data Integration WorkflowsSafe Software
 
99.99% of Your Traces Are (Probably) Trash (SRECon NA 2024).pdf
99.99% of Your Traces  Are (Probably) Trash (SRECon NA 2024).pdf99.99% of Your Traces  Are (Probably) Trash (SRECon NA 2024).pdf
99.99% of Your Traces Are (Probably) Trash (SRECon NA 2024).pdfPaige Cruz
 
9 Steps For Building Winning Founding Team
9 Steps For Building Winning Founding Team9 Steps For Building Winning Founding Team
9 Steps For Building Winning Founding TeamAdam Moalla
 
UiPath Clipboard AI: "A TIME Magazine Best Invention of 2023 Unveiled"
UiPath Clipboard AI: "A TIME Magazine Best Invention of 2023 Unveiled"UiPath Clipboard AI: "A TIME Magazine Best Invention of 2023 Unveiled"
UiPath Clipboard AI: "A TIME Magazine Best Invention of 2023 Unveiled"DianaGray10
 
Bird eye's view on Camunda open source ecosystem
Bird eye's view on Camunda open source ecosystemBird eye's view on Camunda open source ecosystem
Bird eye's view on Camunda open source ecosystemAsko Soukka
 
activity_diagram_combine_v4_20190827.pdfactivity_diagram_combine_v4_20190827.pdf
activity_diagram_combine_v4_20190827.pdfactivity_diagram_combine_v4_20190827.pdfactivity_diagram_combine_v4_20190827.pdfactivity_diagram_combine_v4_20190827.pdf
activity_diagram_combine_v4_20190827.pdfactivity_diagram_combine_v4_20190827.pdfJamie (Taka) Wang
 
Videogame localization & technology_ how to enhance the power of translation.pdf
Videogame localization & technology_ how to enhance the power of translation.pdfVideogame localization & technology_ how to enhance the power of translation.pdf
Videogame localization & technology_ how to enhance the power of translation.pdfinfogdgmi
 
UiPath Community: AI for UiPath Automation Developers
UiPath Community: AI for UiPath Automation DevelopersUiPath Community: AI for UiPath Automation Developers
UiPath Community: AI for UiPath Automation DevelopersUiPathCommunity
 
COMPUTER 10 Lesson 8 - Building a Website
COMPUTER 10 Lesson 8 - Building a WebsiteCOMPUTER 10 Lesson 8 - Building a Website
COMPUTER 10 Lesson 8 - Building a Websitedgelyza
 
UiPath Solutions Management Preview - Northern CA Chapter - March 22.pdf
UiPath Solutions Management Preview - Northern CA Chapter - March 22.pdfUiPath Solutions Management Preview - Northern CA Chapter - March 22.pdf
UiPath Solutions Management Preview - Northern CA Chapter - March 22.pdfDianaGray10
 
100+ ChatGPT Prompts for SEO Optimization
100+ ChatGPT Prompts for SEO Optimization100+ ChatGPT Prompts for SEO Optimization
100+ ChatGPT Prompts for SEO Optimizationarrow10202532yuvraj
 
COMPUTER 10: Lesson 7 - File Storage and Online Collaboration
COMPUTER 10: Lesson 7 - File Storage and Online CollaborationCOMPUTER 10: Lesson 7 - File Storage and Online Collaboration
COMPUTER 10: Lesson 7 - File Storage and Online Collaborationbruanjhuli
 
Machine Learning Model Validation (Aijun Zhang 2024).pdf
Machine Learning Model Validation (Aijun Zhang 2024).pdfMachine Learning Model Validation (Aijun Zhang 2024).pdf
Machine Learning Model Validation (Aijun Zhang 2024).pdfAijun Zhang
 
Salesforce Miami User Group Event - 1st Quarter 2024
Salesforce Miami User Group Event - 1st Quarter 2024Salesforce Miami User Group Event - 1st Quarter 2024
Salesforce Miami User Group Event - 1st Quarter 2024SkyPlanner
 
Artificial Intelligence & SEO Trends for 2024
Artificial Intelligence & SEO Trends for 2024Artificial Intelligence & SEO Trends for 2024
Artificial Intelligence & SEO Trends for 2024D Cloud Solutions
 
Cybersecurity Workshop #1.pptx
Cybersecurity Workshop #1.pptxCybersecurity Workshop #1.pptx
Cybersecurity Workshop #1.pptxGDSC PJATK
 

Recently uploaded (20)

The Kubernetes Gateway API and its role in Cloud Native API Management
The Kubernetes Gateway API and its role in Cloud Native API ManagementThe Kubernetes Gateway API and its role in Cloud Native API Management
The Kubernetes Gateway API and its role in Cloud Native API Management
 
UiPath Studio Web workshop series - Day 6
UiPath Studio Web workshop series - Day 6UiPath Studio Web workshop series - Day 6
UiPath Studio Web workshop series - Day 6
 
Igniting Next Level Productivity with AI-Infused Data Integration Workflows
Igniting Next Level Productivity with AI-Infused Data Integration WorkflowsIgniting Next Level Productivity with AI-Infused Data Integration Workflows
Igniting Next Level Productivity with AI-Infused Data Integration Workflows
 
99.99% of Your Traces Are (Probably) Trash (SRECon NA 2024).pdf
99.99% of Your Traces  Are (Probably) Trash (SRECon NA 2024).pdf99.99% of Your Traces  Are (Probably) Trash (SRECon NA 2024).pdf
99.99% of Your Traces Are (Probably) Trash (SRECon NA 2024).pdf
 
9 Steps For Building Winning Founding Team
9 Steps For Building Winning Founding Team9 Steps For Building Winning Founding Team
9 Steps For Building Winning Founding Team
 
UiPath Clipboard AI: "A TIME Magazine Best Invention of 2023 Unveiled"
UiPath Clipboard AI: "A TIME Magazine Best Invention of 2023 Unveiled"UiPath Clipboard AI: "A TIME Magazine Best Invention of 2023 Unveiled"
UiPath Clipboard AI: "A TIME Magazine Best Invention of 2023 Unveiled"
 
Bird eye's view on Camunda open source ecosystem
Bird eye's view on Camunda open source ecosystemBird eye's view on Camunda open source ecosystem
Bird eye's view on Camunda open source ecosystem
 
activity_diagram_combine_v4_20190827.pdfactivity_diagram_combine_v4_20190827.pdf
activity_diagram_combine_v4_20190827.pdfactivity_diagram_combine_v4_20190827.pdfactivity_diagram_combine_v4_20190827.pdfactivity_diagram_combine_v4_20190827.pdf
activity_diagram_combine_v4_20190827.pdfactivity_diagram_combine_v4_20190827.pdf
 
Videogame localization & technology_ how to enhance the power of translation.pdf
Videogame localization & technology_ how to enhance the power of translation.pdfVideogame localization & technology_ how to enhance the power of translation.pdf
Videogame localization & technology_ how to enhance the power of translation.pdf
 
201610817 - edge part1
201610817 - edge part1201610817 - edge part1
201610817 - edge part1
 
UiPath Community: AI for UiPath Automation Developers
UiPath Community: AI for UiPath Automation DevelopersUiPath Community: AI for UiPath Automation Developers
UiPath Community: AI for UiPath Automation Developers
 
20230104 - machine vision
20230104 - machine vision20230104 - machine vision
20230104 - machine vision
 
COMPUTER 10 Lesson 8 - Building a Website
COMPUTER 10 Lesson 8 - Building a WebsiteCOMPUTER 10 Lesson 8 - Building a Website
COMPUTER 10 Lesson 8 - Building a Website
 
UiPath Solutions Management Preview - Northern CA Chapter - March 22.pdf
UiPath Solutions Management Preview - Northern CA Chapter - March 22.pdfUiPath Solutions Management Preview - Northern CA Chapter - March 22.pdf
UiPath Solutions Management Preview - Northern CA Chapter - March 22.pdf
 
100+ ChatGPT Prompts for SEO Optimization
100+ ChatGPT Prompts for SEO Optimization100+ ChatGPT Prompts for SEO Optimization
100+ ChatGPT Prompts for SEO Optimization
 
COMPUTER 10: Lesson 7 - File Storage and Online Collaboration
COMPUTER 10: Lesson 7 - File Storage and Online CollaborationCOMPUTER 10: Lesson 7 - File Storage and Online Collaboration
COMPUTER 10: Lesson 7 - File Storage and Online Collaboration
 
Machine Learning Model Validation (Aijun Zhang 2024).pdf
Machine Learning Model Validation (Aijun Zhang 2024).pdfMachine Learning Model Validation (Aijun Zhang 2024).pdf
Machine Learning Model Validation (Aijun Zhang 2024).pdf
 
Salesforce Miami User Group Event - 1st Quarter 2024
Salesforce Miami User Group Event - 1st Quarter 2024Salesforce Miami User Group Event - 1st Quarter 2024
Salesforce Miami User Group Event - 1st Quarter 2024
 
Artificial Intelligence & SEO Trends for 2024
Artificial Intelligence & SEO Trends for 2024Artificial Intelligence & SEO Trends for 2024
Artificial Intelligence & SEO Trends for 2024
 
Cybersecurity Workshop #1.pptx
Cybersecurity Workshop #1.pptxCybersecurity Workshop #1.pptx
Cybersecurity Workshop #1.pptx
 

IMCSummit 2015 - Day 2 IT Business Track - The Evolution, Functionality and Benefits of NVDIMMs for Storage and Server Applications

  • 1. PRESENTATION TITLE GOES HERE The Evolution, Functionality and Benefits of NVDIMMs for Storage and Server Applications Arthur Sainio Bob Frey June 30th, 2015
  • 2. Agenda Memory / Storage Hierarchy Evolution of NVDIMMs Taxonomy NVM Trend/Adoption NVM Roadmap NVM Performance How NVDIMMs Work NVDIMM Features Applications Ecosystem and Industry Standardization 2
  • 3. Memory Hierarchy Data-Intensive applications need fast access to storage Large performance gap between main memory and HDD SSDs have narrowed the gap, but a ~1000X gap still exists until a SCM becomes viable for mainstream adoption 3 Performance Gap Adapted from SNIA presentations by Viking, HP NVM NAND Flash Magnetic DRAM Registers Cache CPU Longest Latency, Lowest Cost Shortest Latency, Highest Cost Load/Store (Application Direct Access) Block (Application Indirect Access) NVDIMM NVRAM STT-MRAMPCM Nanotube, Other Acceleration Lowest Latency Highest Latency ReRAM
  • 4. Storage Hierarchy – Persistent Memory 4 Legacy Storage Leading Edge Storage 100 µs 1,000 µs Tier 0 Tier 1 Tier 2 Tier 3 Caching Analytics Database Mail Servers VOD Media Streaming Data Warehouse Archive Content Delivery Backup Surveillance CRM Web Hosting OLTP Indexing In-memory Hot Data Cold Data Access to Host Network Accessible Latency .01 µs 10 µs Memory Channel NVDIMM NVRAM PCIe SNIA presentations by Netlist
  • 5. The Evolution of NVM Origin: to replace battery-backed for NVM cache on storage array controllers NVRAM (Hybrid) PCIe interface, stores in DRAM, backs up into NAND only on a power loss, Supercap based, eliminates battery, maps in MMIO NVDIMM (Hybrid) NVDIMM-N Stores in DRAM, backs up into NAND only on a power loss, Supercap based, eliminates battery NVDIMM-F – maps NAND into memory address space NVDIMM-P - maps NAND and DRAM into memory address space And Beyond! FPGA on NVDIMM Accelerate dynamically changing workloads SCM – Storage Class Memories Next Gen memory technologies (ReRAM, STT-MRAM, PCM) Inherently persistent and removes the supercap module 5Hybrid: Memory subsystem consists of Flash and DRAM
  • 6. JEDEC NVDIMM Taxonomy NVDIMM-F NVDIMM-N NVDIMM-P • One Access Method: direct byte-oriented access to DRAM and a persistence backup/restore function on power fail • Memory mapped DRAM. No system access to Flash • Capacity: DRAM DIMM (8GB, 16GB, 32GB) – uses DRAM and Flash • Latency: DRAM (10’s of nanoseconds) • Energy source for backup • JEDEC type and electrical mechanical definition completed • One Access Method: block-oriented access through a shared command buffer, i.e. a mounted drive. • Memory mapped Flash. DRAM is not system mapped • Capacity: NAND (100’s GB – 1’s TB) - uses NAND flash; the command buffer may be DRAM, SRAM, etc. • Latency: NAND (10’s of microseconds) • JEDEC type definition completed • Two Access Methods: persistent DRAM (–N) and also block-oriented drive access (–F) • Memory mapped Flash and memory mapped DRAM • Supported -> Load/Store, Emulated Block • Capacity: NVM (100’s GB – 1’s TB) – uses DRAM and NAND flash • Latency: NVM (100’s of nanoseconds) • No JEDEC definition yet Single letter designator - combines the media technology (NAND, etc) and the access mechanism (byte, block, etc.) Adapted from SNIA presentations by JEDEC, HP 7
  • 7. NVDIMM Market Trends • NVDIMM market expected to reach $100’s of millions in 3-4 years • Multiple NVDIMM vendors • NVDIMMs are used in server and storage applications (data centers and cloud computing • The market may further take off as the Software ecosystem matures • Industry work underway to develop the load/store stack • The market may follow the PCIe trend starting 2018 with 3.3X growth, 2018-$0.6B to 2021-$5.5B 0 50 100 150 200 250 2014 2015 2016 2017 2018 2019 200GB 400GB 800GB 0 20 40 60 80 100 120 140 160 180 200 2014 2015 2016 2017 2018 2019 8GB 16GB NVDIMM-N Forecast (K/U) NVDIMM-F Forecast (K/U) Source; IHS, Netlist, Mar 2015 $600M $5.5B 2018 8
  • 8. NVM Adoption by System Application 8 DRAM NAND + StorageStorageStorageStorage RAID Controller PCIe Slot NVDIMM Memory Memory Memory Memory CPU StorageStorageStorageStorage NVM PCIe Slot Memory Memory Memory Memory CPU StorageStorageStorageStorage Memory Memory Memory NVDIMM CPU Rack Scale Architecture SNIA presentations by Netlist
  • 9. NVvault® NV-RDIMM (Type-N) Non-JEDEC Standard NVM Roadmap – Enterprise Storage – 2DPC 9 Tick Haswell / Grantley 1.2 V 2133 MT/s 16 GB DDR4 RDIMM Tock Grantley 1.2 V 2400 MT/s 16 GB DDR4 RDIMM DDR4 NVDIMM-N 1.2 V 2400 MT/s 16 GB - RDIMM Tick 1.2 V 2667 MT/s 32 GB DDR4 R/LRDIMM DDR4 NVDIMM-N 1.2 V 2DPC 2667 MT/s 32 GB Type N – R/LRDIMM DDR3 / DDR4 Cost Crossover JEDEC Standard 2H’14 1H’15 2H’15 1H’16 2H’16 1H’17 Product Availability 8/16GB- R 32GB-L DDR4 NVDIMM-N 1.2 V 2133 MT/s 8 GB - RDIMM DDR4 NVDIMM SPD DDR4 Gen2 LRDIMM Raw Cards NV Register Set DDR4 Gen2 LRDIMM Chipset 8Gb DDR4 2400 MT/s DDR4 Gen2 LRDIMM Chipset DDR4 Gen2 LRDIMM Raw Cards Validated Spec Closed Spec Closed Spec Closed Spec Closed DDR4 NVDIMM-F 1.2 V 2400 MT/s 400 GB - RDIMM DDR4 NVDIMM-P 1.2 V 2667 MT/s 16 GB Type N + 400 GB Type F R/LRDIMM NV-LRDIMM (Type-N, -F, -P) JEDEC Standard SNIA presentations by Netlist
  • 10. Netlist 8 GB DDR3 2DPC 1600 Performance Benchmark – NVDIMM- Type N DDR3 - IOPs & Response 10 Performance - IOPs Symmetric Read & Write Performance (1.6M IOPS) Very low Average Response Times (0.02 mS) Bandwidth – MB/s Very Fast Throughput (70GB/s Read / 21GB/s Write- Random) Very Low Average Response Time (1.8 mS R, 5.8 mS W) Response Time (Latency at single OIO or T1Q1) RND 4K 100% Write - .004 mS Average Response Time RND 4K 100% Write - .120 mS Maximum Response Time SNIA presentations by Netlist
  • 11. Form factor is RDIMM/LRDIMM JEDEC module Device model is “SSD on DDR bus”; that is, storage device not a memory device Typically visible to system as block storage device Flash capacity exposed directly to host, as in any SSD 11 NVDIMM(-F) – How it Works NVDIMM-F Controller Flash Host Driver MCU Host Device DDRBus Adapted from SNIA presentations by SanDisk
  • 12. RDIMM/LRDIMM JEDEC module with power backup Data stored into NAND on power loss to achieve persistence Host only addresses the DRAM and has no direct access to the flash NVDIMM controller moves data from DRAM to flash upon power loss or other trigger; can back up portions or all of DRAM upon command When power fails a backup power source provides power to the NVDIMM while DRAM is backed up to Flash MRC (Memory Reference Code) configures NVDIMM controller to move data back from Flash to DRAM when recovery is needed 12 Backup Power NVDIMM MCU NVDIMM(-N) – How it Works
  • 13. Form factor is RDIMM/LRDIMM JEDEC module Three Modes of Operation Storage Mode: “SSD on DDR bus”; Storage device not a memory device, No application SW changes needed, BIOS & Driver required. Memory Mode: DRAM Operation, Volatile NVM Mode Hybrid Mode: Combines NVM and Storage. 13 NVDIMM(-P) – How it Works NVDIMM- Controller Flash Host Driver MCU Host Device DDRBus DRAM
  • 14. NVDIMM (–N) Features NV (Digital) Controller • Facilitates backup/restore functionality • Register access for temp monitoring, ultracapacitor, and flash statistics Standard DDR4 RDIMM Interface 14 SNIA presentations by Netlist DDR3 DDR4 Density Up to 8GB Up to16GB Power (V) 1.5 / 1.35 1.2 Performance 1DPC 1600 MT/s 2DPC 1600 MT/s 1DPC 2400 MT/s 2DPC 2133 MT/s Ultracapacitor Backup • Provides power to move vital data contents from DRAM to flash during a system or power failure event • Eliminates Battery • Charged by NVDIMM via 12V power
  • 15. NVDIMM –N Features DDR3 4GB/8GB Capacity DDR3 1600 DDR3 latency DDR3 Dual voltage 1.35V/1.5V operation Configuration: 1Rx8 RDIMM Interface 12V Aux via a proprietary connection Availability - Now DDR4 4GB/8GB/16GB Capacity DDR4 2400 DDR4 latency DDR4 voltage 1.2V Configuration: 1R/2R RDIMM (2015) & LRDIMM (2016) Interface 12V support via the DDR4 DIMM connector Availability - Now ALL JEDEC Form Factor Support in Intel MRC Follows RDIMM Population Rules SLC or MLC NAND End-to-end Data integrity Backup initiated via SAVE Signal or I2C OS SAVE as part of shutdown SAVE during operation requires system re-initialization of DDR3 register & re-write DDR3 mode register Backup duration 4GB ~ 16GB (Vendor Specific) Health Monitoring Error reporting Predictive failure warnings Backup, restore, power, NAND life status JEDEC or Proprietary 15 SNIA presentations by Netlist
  • 16. NVDIMM Stack Adapted from SNIA presentations by AgigA Tech Platform Hardware NVDIMM BIOS OS CPU/Memory Controller NVDIMM Power Supply SMB MRC + BIOS Modules DIMM Interface Energy Module SAVE Trigger NVDIMM Driver Block Driver Application Byte Addressable Load StoreBlock Mode Software Hardware Kernel Space User Space 21
  • 17. Application Example: HP NVDIMM-N (DDR3) Acceleration – by HP 17 4X the Acceleration inThroughput and Latency Source: SNIA 2/15 Adapted from SNIA presentations by HP
  • 18. Application Example: Storage Bridge Bay (SBB) CPU DDR3 NVM PCIe Card 10GigE NIC DDR3 DDR3 DDR3 CPU NVM PCIe Card 10GigE NIC DDR3 DDR3 DDR3 DDR3 Backplane PCIe PCIe PCIe PCIe Shadow Writes Required for Failover 18 Adapted from SNIA presentations by AgigA Tech
  • 19. SBB: A Simpler/Better/Faster Way CPU DDR3 DDR3 DDR3 NVDIMM DDR3 DDR3 DDR3 NVIDIMM Backplane Non-Transparent Bridge (PCIe) CPU 19 Also a better alternative to Cache-to-Flash implementations: • Separate failure domain • No battery maintenance • System hold-up requirements significantly less severe • 4x write latency performance improvement Adapted from SNIA presentations by AgigA Tech
  • 20. Tier 0 25ns NVDIMM Type 1 (N,F,P) Tier 1 100us PCIe SSD 15us PCIe NVM Tier 2 100us SATA/SAS NVDIMM NVRAM Store Metadata in memory for application acceleration Host Caching: As Cache for direct attached PCIe SSD Fast RAID Computation as block device for distributed storage Check-pointing state for fast sync and restore SSD Mapping Table Persistent RAM Disk As Fast 4K block store Store boot image for fast restore Application Uses – Storage Main Memory & PCIe 20 PersistentVariables Metadata Checkpoint State Host Caching RAMDisk RAID Compute Write buffer SSD Mapping Journaling Logging SNIA presentations by Netlist
  • 21. Advantages of NVDIMMs for Applications Legacy HDD/SSD Solution Persistent data stored in HDD or SSD tiers Slow & unpredictable software stack NVDIMM Solution Persistent data stored in fast DRAM tier Removes software stack from data-path Accelerates SW-Apps ! • DRAM class latency & thru-put for persistent data – 1000X lower latency – 10X+ throughput increase – But, 10X lower capacity vs. SSD • The value is in application acceleration Kernel Module Simplification RAID & Storage Tiering Write Buffers Persistent Caches Kernel Optimization Transaction Logs 21
  • 22. Low Write-Latency Persistent Storage Value Proposition Specifics • Accelerating Write-Latency Datacenter Workloads • Transaction Commits, Logs & Journals 400GB SLC PCIe SSD 8GB NVDIMM Advantage Write Latency (us) >15 <0.1 >150x Read Latency (us) >47 <0.1 >470x Endurance Workload-dependent Unlimited Unlimited PCIe Lanes Consumed 8 0 No PCIe used Power (W) <25 <10 >2x  >100x Lower Write Latency  Unlimited Endurance  Scalable  No PCIe Resources Consumed  Lower Total Cost of Ownership • $/Latency, $/IOPS & IOPS/W Adapted from SNIA presentations by Inphi 28
  • 23. $0M $500M $1,000M $1,500M $2,000M $2,500M $3,000M $3,500M $4,000M CAPEX OPEX TOTAL SATA SSD x2 NV3 x1 $0M $200M $400M $600M $800M $1,000M $1,200M $1,400M $1,600M $1,800M $2,000M CAPEX OPEX TOTAL PCIe SSD x1 NV3 x1 User Case - 5yr TCO Analysis SATA/PCIe vs. NVDIMM 23 100,000 Servers 50,000 Servers 1,000 Servers 98% reduction $2,723M saved 98% reduction $695M saved 98% reduction $3,418M saved 96% reduction $1,771M saved 96% reduction $1,424M saved 96% reduction $347M saved Dell PowerEdge E-26xx CPU DDR3 NVDIMM 1600 MTs Write Intensive App 98% reduction 96% reduction
  • 24. NVDIMM(-N) Ecosystem NVDIMMs & Systems New Applications Open Source Drivers & User API BIOS, MRC Platform Software Mass Deployment Hardware Standardization • System management, Power health • System support H/W trigger (ADR) • Mechanical (power source) • JEDEC NVDIMM Platform Support • Off-the-shelf and OEM platform support for NVDIMM today • System supported H/W trigger (ADR) • Mechanical (power source) Software Standardization • Applications • Linux NVDIMM- aware kernel 4.1 • API’s BIOS Support • NVDIMM-aware BIOS • Intel modifications to MRC to support NVDIMMs • JEDEC NVDIMM I2C command set • JEDEC SPD 30
  • 25. NVDIMM(-N) Standardization JEDEC Hybrid Memory Task Group DDR4 12V Power Pins (1, 145) standardized DDR4 SAVE_n Pin (230) standardized 12V in DDR4 socket will simplify NVDIMM power circuitry and cable routing Under discussion: Standard system interface for NVDIMM i2c register map for NVDIMM-N controller Issued ballot on NVDIMM Controller Event Pin SPD for NVDIMM representation SNIA NVDIMM SIG (Special Interest Group, >20 companies) Formed in 2014 as a SIG of the SNIA Solid State Storage Initiative Communicating existing industry standards, and areas for vendor differentiation Helping technology and solution vendors whose products integrate NVDIMMs to communicate their benefits and value to the greater market Developing vendor-agnostic user perspective case studies, best practices, and vertical industry requirements Standards JEDEC Flow of NVDIMM Adoption and Support BIOS NVDIMM Suppliers Motherboard ODMs. OEMs Platform Integrated Solutions 31
  • 26. Eliminate File System Latency with Memory Mapped Files Application File System Disk Driver Disk Application Persistent Memory Load/Store Memory Mapped Files Traditional New UserKernelHW UserHW Courtesy: 2015 Data Storage Innovation Conference. © 2015 Storage Networking Industry Association. All Rights Reserved.
  • 27. Conventional Block and File Modes Application NVM block capable driver File system Application NVM device NVM device User space Kernel space Native file API NVM.BLOCK mode NVM.FILE modeUse with disk-like NVM NVM.BLOCK Mode  Targeted for file systems and block- aware applications  Atomic writes  Length and alignment granularities  Thin provisioning management NVM.FILE Mode  Targeted for file based apps.  Discovery and use of atomic write features  Discovery of granularities Courtesy: 2015 Data Storage Innovation Conference. © 2015 Storage Networking Industry Association. All Rights Reserved.
  • 28. Persistent Memory Modes Application PM device PM device PM device. . . User space Kernel space MMU MappingsPM-aware file system NVM PM capable driver Load/ store Native file API PM-aware kernel module PM device NVM.PM.VOLUME mode NVM.PM.FILE mode Use with memory-like NVM NVM.PM.VOLUME Mode  Software abstraction to OS components for Persistent Memory (PM) hardware  List of physical address ranges for each PM volume  Thin provisioning management NVM.PM.FILE Mode  Describes the behavior for applications accessing persistent memory Discovery and use of atomic write features  Mapping PM files (or subsets of files) to virtual memory addresses  Syncing portions of PM files to the persistence domain Courtesy: 2015 Data Storage Innovation Conference. © 2015 Storage Networking Industry Association. All Rights Reserved.
  • 29. Building on the Basic PM Model NVM.PM.FILE programming model “surfaces” PM to application Refine API with additional libraries that evolve into language extensions Add compatible functionality to PM file systems Courtesy: 2015 Data Storage Innovation Conference. © 2015 Storage Networking Industry Association. All Rights Reserved.
  • 30. Open Source Sub-initiatives 30 Initiative What will it do? What is the status? 1 NVDIMM Aware Linux Kernel Support for NVDIMM-N modules (arch/x86/kernel/e820.c) Motherboard vendors provide BIOS/MRC changes needed to recognize NVDIMM-N modules. Customers still need either Block or Load/Store Linux driver to enable NVDIMM-N modules. Waiting for the ACPI Spec 6.0 to be published. Availability in Github - April/2015. 2 Block & Load/Store Drivers Vendor agnostic block mode and load store drivers for NVDIMM-N modules. In addition, provides support for DAX (direct access) functionality. The load/store driver will likely come after the block driver into the Linux Kernel. The PMEM initiative is awaiting ACPI related changes to be approved. General Availability in Linux Kernel - TBD 3 DAX Support Direct Access (DAX) support for NVDIMM-N modules in Ext4 Git: https://github.com/01org/prd Path: fs/ext4/* fs/block_dev.c The Ext4 file system DAX support to NVDIMM-N modules eliminates the page cache layer completely. This requires the availability of [2] & [3]. Q2, 2015, in the official 3.20 kernel
  • 31. PRESENTATION TITLE GOES HERE Thank You! 35