jdr@jrx.mlb.semi.harris.com (Jim Ray) (10/29/90)
Has anyone had experience with 5820's connected to HSC70's ( with RA92's )? I have been less than impressed so far with the IO performance and would like to find out if anyone else has had similar experiences with the 5820. What range of IO performance do you get ? Also, we have the CI configured as another network and have been transmitting data to the other 5820's via this mechanism. It appears to be 2x slower than going over the ethernet connection. I would like to find out if we are doing something wrong or if this is normal performance with the 5820. Interresting that the DS5000 IO performance even with scsi drives appears to be faster than the 5820. Again, since we just installed these machines ( about a month ago ), it could be related to setup. -- Jim Ray Harris Semiconductor Internet: jdr@mlb.semi.harris.com PO Box 883 MS 62B-022 Phone: (407) 729-5059 Melbourne, FL 32901
grr@cbmvax.commodore.com (George Robbins) (10/30/90)
In article <1990Oct29.151224.21573@mlb.semi.harris.com> jdr@jrx.mlb.semi.harris.com (Jim Ray) writes: > Has anyone had experience with 5820's connected to HSC70's ( with > RA92's )? > > I have been less than impressed so far with the IO performance and would > like to find out if anyone else has had similar experiences with the 5820. I did some simple tests back when we first got our 5810, and concluded the bottleneck wasn't really the 5810, but rather the HSC50/SDI combination. The SDI peak bandwidth has fallen below that required to effectively support state of the art 24M-Bit/Sec drives. There also seems to be a problem with the buffering (at least in the HSC50), since the HSC statistics seem to show a 10-20% NAK rate. This means that anything beyond an RA81 (or maybe RA82) is is being throttled by the attachment... Performance with the KDM70 should be better, and the board rumored to support the RF series drives in the VAX6000 series should offer performance similar to the DS5000 SCSI at far less cost than the HSC mess. Why DEC can't/won't implement a simple XMI/SCSI disk/tape adapter is a religious matter best resolved by buying a Sun or other "open" system. > I would like to find out if we are doing something wrong or if this > is normal performance with the 5820. Interresting that the DS5000 IO > performance even with scsi drives appears to be faster than the 5820. While we are generally statisfied with our 5810, I'd be hard put to recommend the solution to anyone at the moment. All the DEC rumors seem to point to DECstation and DECsystem 5400 upgrades in the winds, without a peep about the 5800 (except for the DEC ECL MIPS project cancelation, which probably would have fit in the 5800 context). Also, the basic multi-processor support seems to be somewhat up in the air. A cluster of DS5000's with with directly attached SCSI drives seems to offer a hell of a lot better price/performance than the DS5800, plus lower lower long term memory and disk expansion prices. This could change somewhat if DEC decides to deliver a new 5800 CPU board with some MIPS or a memory board using 4M-bit chips, although if FDDI performance is sufficiently impressive, "FDDI/NFS clustering" might be preferable to multi-processor solutions. > Again, since we just installed these machines ( about a month ago ), it could > be related to setup. If you are really distressed with the performance of you 5820, be sure to notify your DEC salespersons of the fact in writing. This should get their attention and put you in a good barginning postion for an alternate configuration, especially if you feel they recommended the DS5820 over more cost effective alternatives or features such as multi-processor capabilities that aren't delivering their full potential. -- George Robbins - now working for, uucp: {uunet|pyramid|rutgers}!cbmvax!grr but no way officially representing: domain: grr@cbmvax.commodore.com Commodore, Engineering Department phone: 215-431-9349 (only by moonlite)
alan@shodha.enet.dec.com ( Alan's Home for Wayward Notes File.) (10/31/90)
In article <1990Oct29.151224.21573@mlb.semi.harris.com>, jdr@jrx.mlb.semi.harris.com (Jim Ray) writes: > Has anyone had experience with 5820's connected to HSC70's ( with > RA92's )? Nope. But is a VAX 8800 with an HSC70 and RA90s close enough. I'd expect them to be similiar. > > I have been less than impressed so far with the IO performance and would > like to find out if anyone else has had similar experiences with the 5820. > > What range of IO performance do you get ? Sequential reading of large files is around 800 KB/sec. The block size of the file system is 8KB and the rotdelay is 4. Writes are about 500 KB/sec. > > Also, we have the CI configured as another network and have been transmitting > data to the other 5820's via this mechanism. It appears to be 2x slower than > going over the ethernet connection. I would expect the CI be faster than the Ethernet, but the other system on my CI is a VAX 8300 and isn't fast enough be interesting. > > I would like to find out if we are doing something wrong or if this > is normal performance with the 5820. Interresting that the DS5000 IO > performance even with scsi drives appears to be faster than the 5820. I've never had my hands on a 58xx to find out what "normal". I would expect it to be comparable with similiarly configued VAX 8800 or VAX 6000-4xx/5xx. A reasonable SCSI disk with a read-ahead cache is nearly always going to be faster than an SDI disk. > > Again, since we just installed these machines ( about a month ago ), it could > be related to setup. If you have spare partition, play around with tunefs(8), rotdelay and maxcontig. > > -- > Jim Ray Harris Semiconductor > Internet: jdr@mlb.semi.harris.com PO Box 883 MS 62B-022 > Phone: (407) 729-5059 Melbourne, FL 32901 -- Alan Rollow alan@nabeth.enet.dec.com
pavlov@canisius.UUCP (Greg Pavlov) (11/01/90)
In article <1990Oct29.151224.21573@mlb.semi.harris.com>, jdr@jrx.mlb.semi.harris.com (Jim Ray) writes: > Has anyone had experience with 5820's connected to HSC70's ( with > RA92's )? I have been less than impressed so far with the IO performance > and would like to find out if anyone else has had similar experiences with > the 5820. > > ..... Interresting that the DS5000 IO > performance even with scsi drives appears to be faster than the 5820. > We do not have an HSC on our 5810, rather KD's to an SA650 (mix of RA70's and RA90's). Our experiences match yours: throughput on a DS5000 appears to be "better" than on a 5810. greg pavlov, fatrf, amherst, ny moretextmoretextmoretext
grr@cbmvax.commodore.com (George Robbins) (11/02/90)
In article <1990Oct29.151224.21573@mlb.semi.harris.com> jdr@jrx.mlb.semi.harris.com (Jim Ray) writes: > Has anyone had experience with 5820's connected to HSC70's ( with > RA92's )? > > I have been less than impressed so far with the IO performance and would > like to find out if anyone else has had similar experiences with the 5820. One thing that it is important to keep in mind is that the DS5800 I/O, especially the CI/HSC combination, is that it is considerably optimized in the direction of aggregate thruput, rather than peak single drive transfer rate. At the moment I have several disk bound jobs running and the effect on interactive response isn't particularly notable, and I could start a couple backups to HSC attached TU78's and they would run "full speed". This is quite different from the kind of I/O perfomance you'd get on a DS3100 or other single SCSI bus system, which might claim several times the single drive peak transfer rate of the CI/HSC/RAXX combination. > Also, we have the CI configured as another network and have been transmitting > data to the other 5820's via this mechanism. It appears to be 2x slower than > going over the ethernet connection. CI networking is an interesting subject. I suspect the CI is somewhat optimized towards block transfers and you won't see any decent throuput with the standard ethernet MTU's, while if you could jack up the MTU and look at something like NFS wich does block transfers, you might get better results. Some of the other high-speed networks have similar problems, they just don't shine when used as pseudo-ethernets. I dunno whether there's any way to make the CI really look like an 80M-Bbit/S network in the Ultrix environment or not. I've heard conflicting stories about whether it's really an effective alternate inter-CPU backbone link and whether using it as a network substantially interferes with CI<->HSC performance. -- George Robbins - now working for, uucp: {uunet|pyramid|rutgers}!cbmvax!grr but no way officially representing: domain: grr@cbmvax.commodore.com Commodore, Engineering Department phone: 215-431-9349 (only by moonlite)
pavlov@canisius.UUCP (Greg Pavlov) (11/02/90)
In article <15458@cbmvax.commodore.com>, grr@cbmvax.commodore.com (George Robbins) writes: > > A cluster of DS5000's with with directly attached SCSI drives seems to > offer a hell of a lot better price/performance than the DS5800, plus > lower lower long term memory and disk expansion prices. This statement (extracted from a much longer posting) is right on the money - literally. We are currently doing just what it says: migrating from a 5810 to a cluster of 4 5000's with SCSI drives, for exactly the reasons given. Our primary application is a mid-to-large database in INGRES (apx. 1GB now and growing at a rate of apx. 10MB per week). One 5000 has replaced the 5810 as a server. When we can keep INGRES from crashing (which has been rare since version 6 arrived...), we see a rough 200-300% overall performance improvement over the 5810. greg pavlov, fstrf, amherst, ny