Ultra-Low Latency SSDs Impact on Overall Energy Efficiency Bryan - - PowerPoint PPT Presentation

▶
ultra low latency ssds impact on overall energy efficiency
SMART_READER_LITE
LIVE PREVIEW

Ultra-Low Latency SSDs Impact on Overall Energy Efficiency Bryan - - PowerPoint PPT Presentation

Dec 15, 2023 121 likes •379 views

Ultra-Low Latency SSDs Impact on Overall Energy Efficiency Bryan Harris and Nihat Altiparmak Computer Science & Eng. Department University of Louisville HotStorage 20 What is Ultra-Low Latency (ULL) data access? Performance Gap

slide-1
SLIDE 1

Ultra-Low Latency SSDs’ Impact on Overall Energy Efficiency

Bryan Harris and Nihat Altiparmak

Computer Science & Eng. Department University of Louisville

HotStorage ’20

slide-2
SLIDE 2

What is Ultra-Low Latency (ULL) data access?

Device Latency Performance Gap

(relative to DRAM)

HDD

∼ 10 ms 100,000×

Traditional SSD

∼ 100 μs 1000×

Ultra-Low Latency SSD

∼ 10 μs 100×

DRAM

∼ 100 ns 1×

2

slide-3
SLIDE 3

What is Optane SSD?

3

Based on 3D Xpoint technology by Intel and Micron, using phase change memory (PCM) Also available as byte addressable DIMM Ideal for applications that require low latency access

slide-4
SLIDE 4

Motivation

Energy characterization

  • Existing studies of Optane SSDs primarily focuses on performance
  • Energy studies look at only individual device or CPU usage

Impact on overall system software

  • ULL disk IO puts increased pressure on system software
  • How does this affect power consumption?

4

slide-5
SLIDE 5

HotStorage ’20

Energy Characterization & Impact

5

slide-6
SLIDE 6

Test devices

6

Technology Interface Model Capacity

Magnetic SATA 3.1 WD Black 7200 rpm 4 TB Flash SATA 3.1 Samsung 850 EVO 1 TB Flash PCIe 3 Samsung 960 EVO 500 GB 3D XPoint PCIe 3 Intel Optane 900P 280 GB

slide-7
SLIDE 7

Experimental setup

7

HOBO plug meter logs power, current, joules, etc., every second.

fio (“Flexible IO tester”)

generates storage IO workloads

Image from ONSET online catalog https://www.onsetcomp.com/products/data-loggers/ux120-018/

slide-8
SLIDE 8

Storage IO workloads

8

Larger data requests More simultaneous requests

slide-9
SLIDE 9

HotStorage ’20

Observations

9

slide-10
SLIDE 10

Observation 1

At idle, Optane uses more power than Flash SSDs

Idle vs. active power consumption

10

slide-11
SLIDE 11

Idle vs. active power consumption

Observation 2

Active power increases with lower latency devices

11

slide-12
SLIDE 12

Read vs. write power consumption

12

HDD (SATA) Flash (SATA) Flash (NVMe) Optane (NVMe)

Observation 3

Newer storage generations have greater differences in power consumption between reads and writes

slide-13
SLIDE 13

Energy proportionality

System power Power normalized to peak power observed Storage performance utilization IOPS normalized to peak IOPS for each device

13

Consuming energy in proportion to the amount of work performed

slide-14
SLIDE 14

Energy proportionality

14

Reads Writes

Observation 4

Newer advancements lead to better energy proportionality

slide-15
SLIDE 15

Observation 5

As latency decreases, pressure on the system software increases, resulting in more overall energy consumption.

Impact on system software

15

Reads Writes

slide-16
SLIDE 16

Impact on system software

16

Effect on CPU Effect on Power

slide-17
SLIDE 17

Overall energy efficiency metrics

17

Performance per unit energy

Bandwidth: Bytes/second = Bytes per joule watt (joule/second) Throughput: IOPS = IOs per joule watt (joule/second)

slide-18
SLIDE 18

Observation 6

Energy efficiency as bytes per joule increases as request size increases.

Impact on overall energy efficiency

18

HDD (SATA) Flash (SATA) Flash (NVMe) Optane (NVMe) Bytes transferred per joule

slide-19
SLIDE 19

Observation 7

Energy efficiency as IOs per joule is coupled to internal parallelism.

Impact on overall energy efficiency

19

Flash NVMe Optane SSD IOs per joule

slide-20
SLIDE 20

Conclusion

Optane SSD has the greatest energy proportionality since its power consumption scales better than previous storage generations based on its range of throughput. Although Optane’s peak power consumption is higher, it yields better energy efficiency as measured in bytes per joule and IOs per joule.

20

slide-21
SLIDE 21

Discussion and future research

Rethink system software for energy efficiency:

  • Polling vs. interrupts?
  • New IO interfaces: io_uring, SPDK, etc.?
  • Simpler mechanisms for blk-mq and NVMe?
  • IO scheduling?
  • Merging, prefetching, buffering, log structuring?

21

slide-22
SLIDE 22

Discussion and future research

Investigate R/W asymmetry:

  • Why do writes use more CPU but fewer interrupts?
  • Do existing design choices get in the way of 3D XPoint?
  • Strategies for hybrid drives for energy efficiency

22

slide-23
SLIDE 23

Discussion and future research

Energy as a whole:

  • Do performance optimizations lose sight of energy efficiency?
  • Compared to CPU and memory, is the impact of the storage subsystem

underestimated?

23

slide-24
SLIDE 24

HotStorage ’20

Thank you!

24

Questions? Please contact us: Bryan Harris bryan.harris.1@louisville.edu Nihat Altiparmak nihat.altiparmak@louisville.edu

slide-25
SLIDE 25

References

[1] Luiz André Barroso and Urs Hölzle. The case for energy-proportional computing. Computer, 40(12):33–37, 2007. https://doi.org/10.1109/MC.2007.443. [2] Kan Wu, Andrea Arpaci-Dusseau, and Remzi Arpaci-Dusseau. Towards an unwritten contract of Intel Optane SSD. In 11th USENIX Workshop on Hot Topics in Storage and File Systems (HotStorage ’19), Renton, WA, July 2019. USENIX Association. https://www.usenix.org/conference/hotstorage19/presentation/wu-kan [3] Sehgal, Priya, Vasily Tarasov, and Erez Zadok. "Evaluating Performance and Energy in File System Server Workloads." FAST 2010.

25