search

ESOS-Lab / VSSIM

Virtual Machine based SSD Simulator
62 stars 50 forks source link

Mixed workloads #2

Open aviadzuc opened 8 years ago

aviadzuc commented 8 years ago

I've been experimenting with VSSIM. I'm seeing some weird behaviour for mixed read/write workload. The throughput I'm getting is much lower than it should be.

I'm running a simple test for a 3GB simulated device, with 8 chip, 1 plane per chip, 4KB pages (I'm attaching my ssd.conf). First I write half of the device sequentially. Than I'm issuing sequential workload for 30 MB starting from sector 0. When I do only SSD_WRITE for every 4KB in that range, I'm geeting about 80MB/s. When I precede each write with SSD_READ of same lba, I'm geeting 10MB/s. Since read latency is much faster this really shouldnt be the case. Worse than that, even if I zero the write latency, I'm still getting the same 10MB/s trhoughput! which indicates a bug.

This is the code I'm running for(i=0;i<8000;i+=SECTORS_PER_PAGE){ SSD_READ(SECTORS_PER_PAGE, i % SECTOR_NB); // preceding read. without this, througput is 8x higher SSD_WRITE(SECTORS_PER_PAGE, i % SECTOR_NB); }

Any idea what might be the problem?

ronniematie commented 8 years ago

Hi there, I am sorry I cannot assist with that porting problem because I am having the same also.

Maybe you can help me by telling me how you managed to run it on Ubuntu with qt3 because i am getting an error while trying:

  1. to make in QEMU, i get this

    make: *\ [subdir-x86_64-softmmu] Error 2

  2. On VSSIM execution i am getting the following error

    Formatting '../../RAMDISK/rd/ssd_hda.img', fmt=qcow size=21474836480 encryption=off ./run.sh: line 10: ./qemu-system-x86_64: No such file or directory

I will greatly appreciate your assistance.

THANK YOU

johnkuriakose12 commented 7 years ago

I am also getting the above issue.