my client got Dell PowerEdge 1950 server with 6GB ram and quadcore cpu installed. There is Flash Media Server 3.52 running on the server on Win2008. External storage Dell MD1000 connected to this server and it has 6 1TB 7200 SATA harddrives running RAID 10. Client got about 1200 connection to streaming at once. We did ask our hosting to set unlimited bandwidth for some time because we had delays during streaming playback at high traffic hours ( evenings ).

I think it has to do with FMS 3.52 limitations or RAID10 IO limitation.

From your experience as person who ever dealt with high traffic tubes how much machines must be set to run 1000 connected at once users and growing each day.

I have no idea where to look, expect buying more harddrives into existing RAID10 array.


  • 1
    Key questions are [a] How much bandwidth do each of these 1200 streams consume [b] What is the absolute external bandwidth limit from your server - 1Gbps, 10Gbps, more? [c] How large are the source files that are used to generate the streams? Those numbers will allow a reasonable estimate of the capacities that are needed.
    – Helvick
    Nov 6, 2009 at 14:42

1 Answer 1


Don't guess if the I/O load is too high. Setup a performance monitor logging every 30 secs on:

Logical Disk - Current Disk Queue length - Drive letter (for each logical drive if there is more than one you care about)
Logical Disk - Disk Read Bytes/sec - Drive letter (for each logical drive)
Logical Disk - Disk Write Bytes/sec - Drive letter (for each logical drive)

Physical Disk - Current Disk Queue length - Drive number (for each hardware controlled RAID array or physical drive if you are doing software RAID)
Physical Disk - Disk Read Bytes/sec - Drive letter (for each array/drive)
Physical Disk - Disk Write Bytes/sec - Drive letter (for each array/drive)

It wouldn't hurt to add a few general purpose counters while you are monitoring so I'd suggest

Processor - % Processor Time - _Total
System - Processor Queue Length

If your % Processor Time is above %50 or if the Queue length stays above 0 on a regular basis you should pay attention to the CPU situation as well.

Back to disk, take account how much used capacity there is on each logical drive.

If your logical drive doesn't hold a large amount of data but the Current Disk Queue Length is high you should move that data to a SSD or a RAID comprised of SSDs. This may only be the boot drive on the server or it may even include your content. For good SSDs think Intel controller (Intel and Kingston) or Indilinx Controller (plenty of brands to choose from). Further if you go this route you need to look at your reads to writes ratio. If your Writes are high you want SLC based SSDs which cost at least twice as much if not more. If your writes are low or especially if the content doesn't change on a regular basis you want MLC based SSDs to save on the cost per GB.

If your logical drive holds an amount that makes the SSD option seem too expensive but you aren't short on space with the existing array you may want to consider getting higher RPM drives. Beware this may be false economy as a couple of good SSDs might handle the load of dozens of rotating disks.

If your drive holds an amount of data that makes you worry about space with the existing number of drives then you can just add more 7200 RPM SATA drives to keep the capacity increasing. Beware this may be counter to the performance issue as it may allow people to add more data and increase the load more than the extra spindles help. If you follow this route you should:

  1. Backup the data
  2. add the extra drives
  3. create a new RAID array using all the disks but not to their maximum capacity.
  4. restore the data.
  5. Continue to monitor performance
  6. repeat as needed

Heck maybe you should do that no matter which of the three types of drives you add but only add them if the performance counter(s) show the need.

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .