ESXi 4.1.0 performs really really bad

Tell me more about how your raid arrays are set up. Also make sure you have the caching memory set up. Note that SATA drives won't give great performance - how does a VM perform if you set it up on the local datastore with the SAS drives?

8x SAS Drives are in a RAID5
8x SATA are two RAID5 because there is a 2TB limit.

Where can I check for the caching memory setup?

Generally in RAID setup utility there is a section for allocating amount of cache memory to allocate to read, amount to allocate to write, read ahead policy, write policy (write through vs write back) etc. How is that configured?

Also you listed read performance - what is the write performance. How are you measuring?

I can press F8 during the start... But there I can only add/remove/view logical drives...

Use the smartstark disk and go into the ACU for this functionality.

I meant HP SmartStart Disk.

Just downloaded it. Going to try tomorrow moring. We already have 00.37 here in central europe...

Just wait a few weeks and it will be daylight savings time on the east coast of US, we'll be 1 more hour back

Okay I checked everything... Was all turned off. So I turned it on. But speed is still very low 32 MB/sec

Well - that was over a 4x improvment - but possibly still a bit low. Just tested and I get around 45-50 MB/sec going to a SATA RAID5 array on a Dell R710 with Perc6i controller under ESXi 4.1. This is with little or no contention on the transfer (it is a lab environment - but the only thing I have close to your specs.). My lab box is also a bit stronger than what you specified in terms of processor and RAM. Perhaps 32 MB is reasonable for SATA RAID5 if you have a moderately loaded LUN you are transferring to or from.

What is your typical load (IOPS) for your SATA array? What are the IOPS when you are testing?

Again, if you configure a vm on the local datastore with the SAS drives what kind of performance do you get?

would a RAID10 be faster? I will do transfer tests on the sas drives tomorrow

RAID 10 is faster for wrties because no parity calculation is required. I wouldn't expect a huge improvement in read performance.

Btw. During the tests... Its the only vm running...

its kinda late now but next time focus on speed and spindles for your vms.  I see you bought the faster better drives for your esxi host but slower, bigger, and cheaper drives for your storage and Vm's.
your esx host can run on almost anything.  It is small, and basically fits into memory.  With esxi you really only need a usb stick.  But drives with raid1 gives you some redundancy.  After initial boot up those drives are hardly used.  Seems like you are using in raid 1 288 gb of space to hold about for no more then about 1gb of data

But for your actual vm's and storage.  More and faster drives are better.  I/O will crush you at some point.  Don't focus on size, but spindles and speed.  Of course make sure you get enough size to actually hold all your data and growth :-)  This is one of those areas where buying the cheaper drives now will cost you 2-3x as much later on.

Okay, I created now a SAS RAID 1+0 Array on the P400 Controller with 8x 10'000 RPM 72GB SAS Drives... Speed with only 1 VM running is 32 MB/s read... that cant be right?!?!

Maybe its your source or destination - what are you reading and writeing from and to?

On the same Raid and VM I am using HD Tach for benchmark

I'm very surprised that the controller write cache was disabled since it's always on by default. You didn't turn the disk write cache on did you? That's very brave if you did, power cut = corrupt data as it's not battery backed.