jeffj@pedsga.UUCP (Jeff Jonas) (11/16/89)
As already posted, IBM has a large disk with two sets of heads. I think the main obstacle is cost. The emphasis for disks are cost capacity size speed power consumption PCs and consumers want low cost and low power consumption. Portables will trade cost for smaller size. Multiple heads will greatly reduce the average seek time, but there's another trick used there too. Say I have a disk with 1000 cylinders. With one set of heads, the head must move the entire of the surface of the disk from edge to edge. Some disks now have multiple heads per arm. Just placing a second head in the middle of the arm means that when the arm is retracted heads are on cylinders 0 and 500, and fully extended they are on cylinders 499 and 999. The arm moves only half the surface of the disk. Logically, the extra set of heads are handled as other tracks, so 10 surfaces with 2 heads per surface appears to be 20 surfaces with 1 head per surface. No need for another linear motor and all the related circuitry. Another consideration is media wear. I heard that in the never ending quest for greater disk capacity, the heads are no longer flying or floating on the disk, but actually riding on a layer of lubricant. Keeping the number of heads to a minimum reduces wear. yes, it would be neat, but there are other solutions that are more cost effective. Jeffrey Jonas jeffj@pecnos.tinton.ccur.com
andrew@alice.UUCP (Andrew Hume) (11/16/89)
imprimis sells a version of their 8" 1.2GB sabre drives that has two independent heads (for a total I/O bandwisth of 6MB/s). its called a "Sabre 2HP". imprimis is also developing a hooning sabre that connects to a xylogics or ciprico controller (which looks like an IPI-3 interface to a host). Each head has an I/O path which means 9x3=27MB/s bandwidth. although in this case, i believe th heads have to be moved together.
terryk@pinocchio (Terence Kelleher) (11/17/89)
>In article <1021@pedsga.UUCP>, jeffj@pedsga (Jeff Jonas) writes: >Multiple heads will greatly reduce the average seek time, but there's another trick used there too. Say I have a disk with 1000 cylinders. With one set of heads, the head must move the entire of the surface of the disk from edge to edge. Some disks now have multiple heads per arm. Just placing a second head in the middle of the arm means that when the arm is retracted heads are on cylinders 0 and 500, and fully extended they are on cylinders 499 and 999. The arm moves only half the surface of the disk. Logically, the extra set of heads are handled as other tracks, so 10 surfaces with 2 heads per surface appears to be 20 surfaces with 1 head per surface. No need for another linear motor and all the related circuitry. This also increases the mass that needs to be moved for seeking and complicates the mechanical makeup of the arms. The NEC 9 inch drives are built this way, but I don't recall that they seek any faster than other drives with single heads per arm. > >Another consideration is media wear. I heard that in the never ending quest for greater disk capacity, the heads are no longer flying or floating on the disk, but actually riding on a layer of lubricant. Keeping the number of heads to a minimum reduces wear. > I don't recall any drives that put the heads directly on the lubricant. I would think this would create a severe heat problem even if the lubricant could protect the media for an extended priod of time. > >Jeffrey Jonas >jeffj@pecnos.tinton.ccur.com -- Terry Kelleher, Encore Computer Phone: 607-798-9602 Internet: terryk@encore.encore.com
terryk@pinocchio (Terence Kelleher) (11/17/89)
In article <10144@alice.UUCP>, andrew@alice (Andrew Hume) writes: > > >imprimis sells a version of their 8" 1.2GB sabre drives that has >two independent heads (for a total I/O bandwisth of 6MB/s). >its called a "Sabre 2HP". The Sabre 2HP does not have 2 sets of heads. It has the same HDA as the standard Sabre 1230. What is has is electronics that allows it to read/write from 2 heads at a time rather than one. It has 2 read/write channels. This gives it higher speed data transfer, but does not improve seek time or latency. > >imprimis is also developing a hooning sabre that connects to a >xylogics or ciprico controller (which looks like an IPI-3 interface >to a host). >Each head has an I/O path which means 9x3=27MB/s bandwidth. although >in this case, i believe th heads have to be moved together. Sounds expensive. Where does the 9 come from? The latest Sabres are 15 head drives. -- Terry Kelleher, Encore Computer Phone: 607-798-9602 Internet: terryk@encore.encore.com
markb@denali.sgi.com (Mark Bradley) (11/18/89)
In article <10144@alice.UUCP>, andrew@alice.UUCP (Andrew Hume) writes: > > > imprimis sells a version of their 8" 1.2GB sabre drives that has > two independent heads (for a total I/O bandwisth of 6MB/s). > its called a "Sabre 2HP". > > imprimis is also developing a hooning sabre that connects to a > xylogics or ciprico controller (which looks like an IPI-3 interface to a host). > Each head has an I/O path which means 9x3=27MB/s bandwidth. although > in this case, i believe th heads have to be moved together. No. The heads are on the same actuator. This is not, therefore, 2 independent sets of heads. Same for their future product. markb -- Mark Bradley "Faster, faster, until the thrill of I/O Subsystems speed overcomes the fear of death." Silicon Graphics Computer Systems Mountain View, CA 94039-7311 ---Hunter S. Thompson ******************************************************************************** * Disclaimer: Anything I say is my opinion. If someone else wants to use it, * * it will cost... * ********************************************************************************
andrew@alice.UUCP (Andrew Hume) (11/19/89)
sorry to have introduced any confusion about headedness. in the absence of manuals, i went on what my imprimis sales rep said. i stand by the figure of 27MB/s for the drive under development. although i do not understand where the nine comes from either. perhaps just nine of the heads have i/o paths; the number nine might be due to bandwidth considerations (how fast can IPI go?) or from total head mass limits. i still believe this disk to be cheapish, although with the controller it might be as expensive as an ra81 ($15K).