stevec@otl.SanDiego.NCR.COM (Steve Chaput) (01/17/90)
I have a question concerning Sybase configurations. In a hypothetical situation, we have a Sun dedicated as a Sybase 4.0 SQL Server on a LAN. Assume that we have tuned our server for optimum performance. However the client application accessing the server requires greater performance. The client application accesses ATM machines and performs single row updates, inserts, and fetches on a very large table (350MB). We want to dedicate another identical Sun to share the server requests to increase client performance. We are willing to modify the client application and partition the data between machines. Replication of the data is undesirable. How can we decrease the time it takes the client to process n transactions utilizing the additional machine ? How much of a change could be expected ? How would other database vendors (Ingres, Oracle, Informix...) handle this situation.
tim@binky.sybase.com (Tim Wood) (01/18/90)
In article <2358@ncr-sd.SanDiego.NCR.COM> stevec@otl.SanDiego.NCR.COM (Steve Chaput) writes: > > >I have a question concerning Sybase configurations. In a >hypothetical situation, we have a Sun dedicated as a Sybase >4.0 SQL Server on a LAN. Assume that we have tuned our server >for optimum performance. However the client application >accessing the server requires greater performance. The client >application accesses ATM machines and performs single row >updates, inserts, and fetches on a very large table (350MB). >We want to dedicate another identical Sun to share the server >requests to increase client performance. We are willing to >modify the client application and partition the data between >machines. Replication of the data is undesirable. > >How can we decrease the time it takes the client to process n >transactions utilizing the additional machine ? How much of a >change could be expected ? > >How would other database vendors (Ingres, Oracle, Informix...) >handle this situation. I think a worthwhile approach would be to increase the concurrency on the client side. If your incoming requests are serialized through the client, the client will probably be saturated before the server. In an ATM network, the organization for maximum concurrency would be to have each ATM running its (client) side of the transaction. That way, almost all the concurrency arbitration happens in the server, which is optimized for it. I say "almost" to allow for network latency. That can be offset by connecting the server to several networks and allocating the ATMs among the networks. Allocation could be static, or dynamic and decided for each connection. For a sufficiently intelligent single network, it will present a virtual "single network" interface to each end and do good congestion control. -TW --- Sybase, Inc. / 6475 Christie Ave. / Emeryville, CA / 94608 415-596-3500 tim@sybase.com {pacbell,pyramid,sun,{uunet,ucbvax}!mtxinu}!sybase!tim This message is solely my personal opinion. It is not a representation of Sybase, Inc. OK.