Answered 3PAR versus EVA

[gerie]

I have a progress database10.1C from about 95 gb.
The data was located on an EVA storrage.
Since 2 months the storage is changed from Eva to 3PAR.
I was told this system would have a better performance.
My online back-up's had a running time from about 1,5 hours and then it was finished.
But after the change from EVA to 3PAR the running time from an online back-up is between 3,5 and 7 hours.

Imo this is not acceptable. My system administrator keeps telling me that I just can't work with full back-ups anymore, but I have to use transactionlogging now.
I'm wondering what your thoughts are about this issue.

 

[Rob Fitzpatrick]

For starters, I'm not familiar with EVA or 3PAR. That said, if my SA came to me and said "guess what, I'm going to put your DBs on new storage with 2X to 5X slower performance", I'd tell him to keep his hands off my DBs. If online backup time has increased that much, what is this new storage subsystem doing to your application performance?

You need to stand your ground. SAs are supposed to keep systems up and running and maintain them. They are not supposed to tell DBAs what their backup strategy should be.

You should explain what he meant by the "transaction logging" comment. Did he mean that you're not using AI now and you should in future? If that's the case then yes you should be using AI in production. But that doesn't obviate the need for daily backups, and it doesn't make a drastic reduction in backup performance any more acceptable. AI is key to recovery scenarios, but backups are required in lots of maintenance activities outside of disaster recovery where you're not going to use AI.

If you still have access to the EVA storage, run application tests and measure the effect of moving to the new storage on application performance. Build your case and back it up with numbers, then escalate. Sounds like you have a turf war ahead of you, and you need to put the SA on the hot seat and make him answer for this "downgrade".

 

[gerie]

I do use AI for disaster recovery. I had all my systems and distaster recovery enviroment working perfectly, until now.
So I try to keep stand and tell him that he just has to make sure my systems are running conform acceptation.

I'm already escalating this problem, but management seems to have problems filtering what's true or false.

 

[cj_brandt]

The performance probably isn't so much about the vendor - 3par or EVA, but how the storage is configured.
It usually comes down to RAID level and cache.

Instead of talking about backup times, how about a corrupt index on a frequently used table and have to run idxbuild to get the system working again. How much longer is the outage going to last because of the new storage ?

The old progress ATM benchmark is a good way to an idea of the performance of a new system.

Another option is to create a test database with 16kb bi blocksize and 32mb bi cluster. Then use bigrow to create 4 new clusters and time how long it takes to add the clusters on the 2 storage systems.

 

[gerie]

I scheduled a dump and load action last weekend, but I blow it off, I fear that the re-indexing in the current situation will take me a week........
Recovering the database on another system takes less time then the creation of a backup on my production server.
I can't influence the configuration of the storrage. I now asked for monitoring logs every 5 minutes with a messurement of I/O, CPU and memory usage. And I hope something will come up as an explenation.(I'm just desperatly searching.)
On the other hand, I must assume, the SA's already monitored the server, so I probably won't find anything.

 

[TomBascom]

SANs do not exist to support databases. Databases get put on them for all the wrong reasons. And are therefore almost always misconfigured. NAS (Netapp) is even worse.

Changes to storage subsystems are pretty much never about making things faster. They are almost always about making things "easier" in some way for the storage administrator.

 

[gerie]

Oh well, with the dedicated space I had on the former configuration, it worked fine. It's only since I migrated to 3PAR that I encounter problems.
Oh wait there is an other detail. I got some more diskspace with the transfer, so that is the other difference. :-(

 

[urgent]

Hi Gerie, i may not be able to give you a technical answer but it seems like Rob, Cj & Tom input contain great amount of information's.

My advice to you is, it sounds like you were given a "promise" that the system\backup will perform faster and that was far from truth. therefore from now on i suggest you run full (performance, functional & stress) test before moving on to a new system or any type of technology that might have impact on your db's....