LVM vs onspaces and Ideal Chunk sizing

Greetings All, Is there a benefit to using: LVM / Harware based partitioning tool versus onspaces using offsets to create raw device chunks for IDS? For example, I have a 2 TB array (mirrored and striped) that I want to carve up into 400 5 GB chunks to assemble a set of dbspaces. I want to keep the chunk sizes/# somewhat reasonable to allow parallel page cleaner operations to occur, minimizing potential checkpoint stall. Is there an advantage to creating 400 raw devices with the LVM versus a few raw devices (say 10) and then using onspaces to identify offsets into each of these 10 larger partitions? Seems like 6 of one / half dozen of the other (perhaps more work for the SA). Also, is there an 'ideal' general chunk size that isn't so big it affects flush duration at checkpoint? Is 5 GB too small or can I get away with a bigger size? I'm trying to balance the size of the chunks versus the number to keep I/O in balance. I know, move to 11.x and this isn;t so much of a problem (working on it). This system is primarily OLTP with some DSS type reporting. IDS 10.00.FC5 Solaris 10 Sun V490 - 4-way 1.5 Ghz US IV+, 16 GB RAM HDS AMS 1000 All insight appreciated,

<the_omegamon@yahoo.com> wrote in message news:222c556b-a712-488b-872c-00bb585952d7@34g2000hsf.googlegroups.com... > Greetings All, > > Is there a benefit to using: > > LVM / Harware based partitioning tool > > versus > > onspaces using offsets > > to create raw device chunks for IDS? > > For example, I have a 2 TB array (mirrored and striped) > that I want to carve up into 400 5 GB chunks to assemble > a set of dbspaces. I want to keep the chunk sizes/# > somewhat reasonable to allow parallel page cleaner > operations to occur, minimizing potential checkpoint > stall. > > Is there an advantage to creating 400 raw devices with > the LVM versus a few raw devices (say 10) and then using > onspaces to identify offsets into each of these 10 > larger partitions? Seems like 6 of one / half dozen of > the other (perhaps more work for the SA). Do you mean LVM? Or Solaris Volume Manager? Ordinarily I'd use LVM (if it were available on Solaris), simply because I think it would make your onstat -d output easier to interpret. But I find Solaris Volume Manager, although functionally adequate, definicient in tools to enable you to keep a track of what is where, and particularly at a glance how much is used and how much is left. So, more marginal, but probably I'd still avoid the offsets.