Progress in the Virtualized Storage World

[doom1701]

There's one thread here discussing VMWare, so it doesn't appear that virtualization is that big of a topic. I'm curious, though, if anyone is using SANs for their disk storage, even if your Progress box is physical.

I've always been told that you should have your BI, AI, and DB on separate physical disks. This worked fine in the physical world; I've currently got a BI mirror, an AI mirror, and a DB RAID 10 array. I've even got the DB on a separate disk controller from everything else.

So far, when I've been playing with virtual progress boxes (our Test SX server is now a virtual server), I've been following the same model with virtual disks. But the disks are really just connections to one SAN (with one large RAID 5 array on the physical disks--I know, let's not get into the RAID 5 discussion), so there's really no point. There's no additional redundancy or speed.

Someday I want to see us running everything virtualized, with storage on a SAN rather than physical storage on servers. But to maintain that same recommended configuration, I'd have to have physical disks in their own arrays on the SAN dedicated to BI, AI, and DB--at which point there's no reason to have them in the SAN, when I could connect them directly just as easily. Dedicating physical disks on a SAN pretty much removes all of the benefits.

So I'm curious if anyone is using a SAN for their Progress storage, and, if so, how are you setup?

 

[TomBascom]

People use SANs for Progress all the time.

In my world it is, by far, the default configuration and has been for a long time.

Stand-alone BI disks haven't been significantly important for anything other than benchmarks since v6.

Stand-alone AI disks are a matter of recoverability. If you lose the underlying disks and both the AI and the DB were on the same disks you can no longer roll-forward. Believe it or not SANs do fail. Catastrophically even. So this remains important. It is also very important to archive the ai logs to an off site location early and often. I like to ftp them elsewhere right after making a local copy of them.

RAID5 is still poison for performance but it has become more bearable for many people as the size of the RAM caches has increased dramatically.

RAID5 also becomes less reliable as it ages -- double disk failures turn out to be much more common than the early analysis thought.

There are some advantages to have distinctly identifiable filesystems that map to psuedo devices on the SAN. If you arrange them well and your OS and SAN tools support your efforts it makes it a lot easier to keep track of where your IO activity is.

 

[tamhas]

I.e.

In some cases costs have changed, allowing us to do things we couldn't consider before.

In some cases device performance has improved overall such that some distinctions between alternate configurations are less important than they used to be.

In some cases, efficiency and utilization has improved, again lessening the importance of the differences.

But, in the end, physics is physics and no amount of wishing or market hype will make it not so.