Disk and Thermal Emulation in Data Center with RAMP

1
Disk and Thermal Emulation in Data Center with
RAMP
2
Introduction

• Problems attacked in this work
• Emulate computer system in data center using RAMP
  (Research Accelerator for MultiProcessing)
• FPGA based HW emulation
• Softcore processors run real workload
• Timing accurate and great scalability
• RAMP targets processor emulation, but it can do
  more!
• Local storage emulation (Disk)
• System temperature emulation
• 50 MHz vs 2 GHz

3

• Introduction Background
• Methodology
• Experiments, Demos and Results
• Limitations and Future Work

4
Background

• Disk is an essential part for data center
• Local physical storage
• Need to understand its impact on system
  performance and power
• Why disk emulation is hard?
• Provide real physical storage to run something
• Timing accurate (VM only have a functional
  module)
• Accurate input for analytical Power estimation
  model
• Correctly reflects to physical disk
  specifications (e.g. 7200RPM)
• In the context of real workload
• Temperature is a critical issue
• Cooling, reliability
• How the workload will affect the temperature in
  data center is an interesting topic

5
Methodology

• Target system (FPGA) 32-bit Leon3 with Linux 2.6
  kernel, 50MHz, 90 MHz DDR memory
• Emulating IDE disk with Ethernet based network
  storage (ATA over Ethernet) DiskSim
• AoE Encapsulate IDE command in Ethernet packet
• DiskSim widely used disk simulator (provide
  access timing based on disk specification)
• Thermal emulation is done by Mercury suite
  (ASPLOS 06)
• Sample CPU/disk activities periodically and send
  to a central emulator
• Emulator takes system configuration and predict
  temperature based on Newtons laws of cooling
• Time dilation makes target looks faster
• Reprogram HW timer to make jiffies longer in
  terms of wall clock
• Slow down memory accordingly, when speeding up
  processor

6
Putting everything together
7
Experiments and Demos

• Target system model for thermal emulation
• Physical layout from Dell PowerEdge 2850
• 3 GHz Xeon, 10K RPM SCSI
• Emulated disk model (validated disk model in
  Disksim)
• Seagate Cheetah 9LP
• 10K RPM, 5 ms avg seek time
• Several programs run in target system with
  various time dilation factors
• Dhrystone CPU intensive benchmark
• Postmark A file system benchmark (disk
  intensive)
• Unix command with pipe (both disk and CPU
  intensive)
• cat alargefile grep a search pattern gt
  searchresultfile
• 100 MB file size
• Emulation output
• Performance statistics
• System temperature

8
Dhrystone result (w/o memory TD)
How close to a 3 GHz x86 8000 Dhrystone MIPS?
Cache effect?
9
Dhrystone w. Memory TD
Keep the memory speed constant in target system
- 90 MHz DDR DRAM in target (50MHz to 2GHz)
- Dilation with a constant cycles
10
Postmark file system benchmark

• Speed-up factor is larger than TDF
• How close to modern SATA disk? Twice throughput
  if run the same benchmark.

11
Disk emulation performance

• Larger TDF offset the network overhead
• Overall emulated disk time still a little longer
  than simulated timing (2.8 ms)
• Time accuracy in OS becomes the problem.

12
Emulated disk R/W time in target

• Pretty deterministic result with different TDF

13
CPU Temperature Emulation
50 MHz
250 MHz
500 MHz
1 GHz
2 GHz
14
Disk Temperature Emulation
50 MHz
250 MHz
500 MHz
1 GHz
2 GHz
15
Limitations

• AoE limits the maximum number of RW sectors to 2
  !(Ethernet packet limitation)
• Naïve memory dilation (constant delay)
• Problem with different time dilation factors
• Small factors network delay dominates
• Large factors time measurement accuracy in OS
  affects the accuracy

16
Future Work after the semester

• Better time dilation model
• Emulate real-life disk controller (plug Disk
  Emulator to low-level device driver)
• IDE controller Intel PIIX4/ICH5, (with DMA
  support), AHCI (SATA controller)
• SCSI