Hi Andy,
> What are assuming your backing store will be?
That's open right now. Previous systems have gone for direct attached
u320 scsi. We're considering staying RAM resident until end of day and
only hitting disk then. For redundancy we can stream to a secondary (and
beyond) from the primary head machine. With full duplex networking we
can stream that out without affecting the inbound.
This is complicated as people are also interested in 24/7 data streams
as well as the more traditional day's worth.
> from other streams (see below) poses additional burdens. You may also
> need to use more than one server here.The highest single-server op rate
> we've ever measured was about 90,000/sec (two CPU Sun v20z); using a
> DRAM storage model.
This is pretty much what we expect to do. We could load balance at layer
3 to multiple writers, but it really doesn't seem worth the trouble
given that we want to write the data as sequentially as possible.
> requests. Now, what is the problem is the "other" non-high transaction
> rate work that would be going on at the same time. We've found that
> mixing such workloads on a single server or device hurts one or the
> other and many times both. I don't think you can expect a high
> transaction rate if the data holder has to also support "bulk" requests.
Yes, this is a good point. One approach I'm looking at is having
secondary copies of the data on multiple servers be the ones that cope
with the bulk requests. Then, when they are finished with that they need
to 'catch up' with the data the primary has since written.
niall
|