Anyone pegged the SATA 1.5gb/s drives yet?

[AJ Bertelson]

Tell me I am not crazy.

I am looking at switching over to HD and probably bring along the two PVR 150's I have. The proposed setup is the following

HD-PVR
HPVR 2250 or HD homerun
PVR-150
(2) 750gb drives sata 1.5gb/s seagate 7200.10

I really doubt that I would hit the bandwidth of these drives going in as the max bandwidth this setup could produce for the PVR's is about 45mb/s correct as its 3x15mb/s for the HD recording, but I figured I would ask.

[bastian74]

The bandwidth is more about how many drives you have on the same bus. You'll not hit the limit with one drive, and its not likely you'll ever hammer more than 1 or two drives at a time.

You'll start seeing benefits of sata3 when you have 4 or more drives in a raid array. Less than 3 drives you wont even max out sata1.5

[AJ Bertelson]

good to hear, I am also considering raid this array but the try to find a good raid card that isn't over $100 isn't easy.

[Taddeusz]

I think it's more of a problem of hitting the limit of seek latency on a drive rather than the maximum bandwidth of the interface. I've run into issues where for whatever reason I have two recordings running off a single drive plus ShowAnalyzer is cranking on both of them. Then I start watching one of them on my HD100 or client and get start/stop playback. I've tried various ways of fixing the problem but I've decided it all comes down to hitting the seek latency on the drive. Just too many thing going on at once and the drive can't keep up with all of them at the same time.

[CollinR]

Agreed, those figures are basically for the cables the actual read/write and head movements are nowhere near that fast.

Taddeusz, you might try a software called ultimate defrag, it will allow you to do whats called a consolidated defrag. So in addition to defragging it will line you stuff up on the outside (or inside) of the disk moving all free space to the opposite end. This will radically effect your head seek times.

[Taddeusz]

Originally Posted by CollinR

Agreed, those figures are basically for the cables the actual read/write and head movements are nowhere near that fast.

Taddeusz, you might try a software called ultimate defrag, it will allow you to do whats called a consolidated defrag. So in addition to defragging it will line you stuff up on the outside (or inside) of the disk moving all free space to the opposite end. This will radically effect your head seek times.

Seek latency was actually never a problem for me until I started recording HD and got ShowAnalyzer set up. And I think I have that problem pretty much fixed. I have DirMon set to run as many copies of ShowAnalyzer as I have tuners. That way each recording is being processed in real-time and theoretically shouldn't be needing to thrash the disk as hard. Particularly since some of the data is going to be cached.

IMHO, Sage needs some kind of load balancing routine to properly spread writes across multiple recording directories which should really be separate drives. I'm not sure how they have it working now but it fails to properly take into account load on each individual recording source and I believe relies upon recoverable space to determine where programs will be recorded.

I've had nights where all three tuners were banging away at the same hard drive and as a result I couldn't watch anything.

[alfi33]

Originally Posted by Taddeusz

And I think I have that problem pretty much fixed. I have DirMon set to run as many copies of ShowAnalyzer as I have tuners. That way each recording is being processed in real-time and theoretically shouldn't be needing to thrash the disk as hard. Particularly since some of the data is going to be cached.

Taddeusz, can you explain this a little more? I'm not sure I understand.

If I try to watch something I previously recorded while I have my both my tuners recording HD shows and a session of Comskip running, my recording hard drive gets hammered and I get very stuttery playback. Are you saying that if I ran another session of Comskip that it might help?

[Taddeusz]

Originally Posted by alfi33

Taddeusz, can you explain this a little more? I'm not sure I understand.

If try to watch something I previously recorded while I have my both my tuners recording HD shows and a session of Comskip running, my recording hard drive gets hammered and I get very stuttery playback. Are you saying that if I ran another session of Comskip that it might help?

Essentially, that was my idea. Since ShowAnalyzer is very disk intensive running full bore on already recorded programs I decided that the best way to limit it was to have it run basically in real-time as they're being recorded. That way it limits the impact it has on disk.

[alfi33]

Originally Posted by Taddeusz

Essentially, that was my idea. Since ShowAnalyzer is very disk intensive running full bore on already recorded programs I decided that the best way to limit it was to have it run basically in real-time as they're being recorded. That way it limits the impact it has on disk.

Very interesting. Seems a little counter-intuitive to me. I would have thought that another session would make the drive have to work harder....but what do I know! I'll give it a try! Thanks for the idea.

[Taddeusz]

Originally Posted by alfi33

Very interesting. Seems a little counter-intuitive to me. I would have thought that another session would make the drive have to work harder....but what do I know! I'll give it a try! Thanks for the idea.

One would think so but because it's working on recordings that are concurrently being recorded it would limit it to real-time. Which theoretically would be dealing with a lot of cached data as long as you have the RAM.

[toricred]

Originally Posted by bastian74

The bandwidth is more about how many drives you have on the same bus. You'll not hit the limit with one drive, and its not likely you'll ever hammer more than 1 or two drives at a time.

You'll start seeing benefits of sata3 when you have 4 or more drives in a raid array. Less than 3 drives you wont even max out sata1.5

I thought the point of SATA was that each one had it's own bus. That's why there's no setting for Master or Slave.

[kbyrd]

Originally Posted by toricred

I thought the point of SATA was that each one had it's own bus. That's why there's no setting for Master or Slave.

For internal drives, this is usually true. However, some sata chipsets support port multipliers (usually for eSATA, "external SATA"). In that case you have an external SATA cable going to a drive enclosure which then contains 4 or so drives. With this, you could hit 1.5Gbps or maybe 3Gbps on that single eSATA connection.

[waynedunham]

Originally Posted by Taddeusz

Essentially, that was my idea. Since ShowAnalyzer is very disk intensive running full bore on already recorded programs I decided that the best way to limit it was to have it run basically in real-time as they're being recorded. That way it limits the impact it has on disk.

Interesting theory. I may have to give that a try. I know that back when the server was transcoding for my MVP in the bedroom I had to put severe limits on SA in order to assure that Sage had enough horsepower left to run the transcode process on my puny P-4 3.0ghz.
So I set up SA to wait for a file to be done recording, only run one instance, and severely limited it's resources. Right now I've got it just about right. It can mark commercials faster than viewing in SD programs and is just slightly faster than viewing for most/all HD recordings.
After getting my HD-100's and elminatiing the need for transcoding HD recordings I never revisited the issue.

I have a P-4 3.0ghz 1 gig ram, 2 SATA 1 drives (each on it's own SATA controller). I have no troubles recording 2 HD streams (and that's before the new HDHR firmware that only records the actual channel/sub instead of the whole channel), and 4 SD streams (2 PVR-500's) recording at 2gigs per hour.
The drive(s) can sound a bit "busy" when all the tuners are recording and SA is running, but other than a few 1-2 second spinning icons I have no troubles whatsoever. And even then those spinning icons usually only show up when its switching the tuners from one recording to the next and/or when SA fires off a new process for the next recording. And that only lasts for less than a minute tops. (total duration of the "lag", spinning icon only lasts for 1-2 seconds)

Anyway, I may have to experiment with your method. As you said, on the surface it seems counter intuitive, but when you really think about it it might just make sense.