Both of our StarWind Servers failed with 10 minutes of each other. Now that both servers are back up, everything is out of sync and nothing sees any targets. If I click Fast Synchronization, I get the message "Parter is not synchronized". If I click Full Synchronization, it says "Are ...

These tests were run on hybrid SSD and 15K systems. Each of the servers has 2 LSI 24 drive 620J JBOD's connected to a 9280-8E with CacheCade. Each server has a 40 drive RAID 60 of 15K Savvio's with 8 Intel X25-E 32GB SSD's for CacheCade reads. We're ramping up to upgrade our entire order entry, bill...

I did some more testing. 2008 R2 MPIO performance is outstanding with this version. These tests were run from an HP BL465c G7 in a c7000 chasis with 2 Virtual Connect Flex 10's. It has two 8Gb iSCSI HBA's set up. This is going to be one of our 5 new database servers. You have to love having no hard ...

It hasn't done it now for 2 days. I'll run some heavy tests against it to see if I can get it to happen again.

I managed to get 4 of 4 multipathing paths to our 2 HA StarWind 5.7 servers connected to 3 Xen 5.6SP2 servers thanks to the Xen and StarWind HA Survival Guide R2 that Robert sent me. I've started setting up a couple master VM's and have run into an issue. Whenever I start doing some heavy writes to ...

I agree, with the cache being dynamic, specifying how much to use for read and how much for write doesn't make sense. I think I'll go with 2 of the 4 port 10GbE x16 NICs in each server. Each 4 port will probably have the same real world aggregate throughput as a 6 port. If I need to, I can always th...

That all sounds great. By read/write ratios I was referring to how some RAID controllers allow you to allocate certain percentages of the cache to read and write specifically. With dynamic control of the cache, I don't think it would be necessary. Dynamic with a good algorithm behind it would probab...

I was using SQLIO because the majority of my RAID tuning has been for SQL, and those are the numbers I'm used to looking at. Makes it easier for me to compare to other storage systems I've worked with. I'll run some more tests with the other tools for you, and to start collecting other stats for mys...

Here are some numbers for random writes to the single 4GB cache HA target. :shock: C:\Program Files (x86)\SQLIO>sqlio -kW -BN -o16 -frandom -b64 -LS -Fparam.txt sqlio v1.5.SG using system counter for latency timings, 3579545 counts per second parameter file used: param.txt file h:\testfile.dat with ...

Definitely want to make sure that you crank the IOPS down from the default 1000 for ESX round robin. Here are the same original 3 tests to the target with 4GB cache with IOPS set to 1. C:\Program Files (x86)\SQLIO>sqlio -kW -BN -o8 -fsequential -b64 -LS -Fparam.txt sqlio v1.5.SG using system counter...

Wow, look at this. I was just fooling around and did a RANDOM 64K write to 3 2GB files with 1 on each of the test targets. Over 6600 IOPS and 400 MB/s with an average of 6 ms latency! C:\Program Files (x86)\SQLIO>sqlio -kW -BN -o16 -frandom -b64 -LS -Fparam.txt sqlio v1.5.SG using system counter for...

Here are the same set of tests but with 4GB write back cache C:\Program Files (x86)\SQLIO>sqlio -kW -BN -o8 -fsequential -b64 -LS -Fparam.txt sqlio v1.5.SG using system counter for latency timings, 3579545 counts per second parameter file used: param.txt file h:\testfile.dat with 8 threads (0-7) usi...

Ok, you guys definitely made some changes to the writes with no cache. These are a VMware 4 VCPU 3.33GHZ with 16GB RAM to a 250GB HA target set to no StarWind caching. Round robin with default 1000 IOPS. C:\Program Files (x86)\SQLIO>sqlio -kW -BN -o8 -fsequential -b64 -LS -Fparam.txt sqlio v1.5.SG u...

No problem. I'll set it up and test it.

They were so bad before when I first tested them that I haven't bothered setting up any targets without write back caching. That's why there is 72GB RAM in each StarWind box.

Bohdan (staff) wrote:And what are the results for HA target with no StarWind caching?