| Title: | VAX DBMS |
| Notice: | THIS NOTESFILE IS NOT A FORMAL SUPPORT CHANNEL |
| Moderator: | SCARY::CHARLAND |
| Created: | Thu Feb 20 1986 |
| Last Modified: | Tue Jun 03 1997 |
| Last Successful Update: | Fri Jun 06 1997 |
| Number of topics: | 2642 |
| Total number of notes: | 11044 |
I was talking to customer who is considering moving from VAXes to
Alphas. I recommended going for two Alphas, not clustered, with hot
standby, using one for production and the standby for reports.
Their average transaction duration is 1s, IO stall time per transaction
9/100s, lock stall time per transaction 1/100s. They want to reduce the
average transaction duration. Availability is a high priority for
them. They have a cluster of 4 VAXes and have recently added a Gbyte of
memory to each. They run production on 3 and reports on the 4th which
is also the lock master. Each system has several CPUs. KERNEL and
INTERRUPT use more CPU than EXEC which uses more than USER mode. There
is a fair amount of idle time.
I know there isn't much detail. My visit was cut short, and this wasn't
its main purpose, but comments on my suggestion would be welcome.
Peter
| T.R | Title | User | Personal Name | Date | Lines |
|---|---|---|---|---|---|
| 2632.1 | sounds good to me | HOTRDB::LASTOVICA | Use a fork Luke! | Tue May 20 1997 12:45 | 6 |
certainly from a performance standpoint, one larger system will be
'better' than several smaller ones (less cluster traffic, all locks are
local, etc.). Once digital proposes the right size of a machine for
them (to replace the vaxes), the database config 'should' be pretty
easy. Plan on running a current version of VMS (7.0 or later) and some
of the older balsetcnt issues of large machines should be resolved.
| |||||