According to the user guide the WAN Accelerator needs "HDD: minimum 50 GB for WAN accelerator cache". But it looks like it needs some amount above the size of the cache because I'm getting failures due to lack of disk space. I have a 100GB volume and I set the cache to 99GB. I now have 96KB free on that drive and every VM is failing with this message:
11/26/2013 5:50:18 AM :: Processing 'EX-06' Error: An existing connection was forcibly closed by the remote host
--tr:Cannot write data to the socket. Data size: [1048576].
--tr:Failed to send block ( offset: '7339864', data size: '1048552', index: '7')
Send file 'E:\VeeamWAN/Send/5d98b576-7829-4c28-b389-7aa04b34c86c\2efa8ed7-44b7-40dc-95d4-f4281be253d1_25db8072-a75d-4272-9557-6237c1a9bea4_2001_088b3887-7b82-49bc-9b04-40ddb7d672f2.0.gz' from offset 0 failed.
--tr:Send file failed.
--tr:Wan connection failed to run command.
--tr:Sen
Is there a formula I can use to calculate how much free space I should add to the drive?
You definitely need more space on the drive than just the global cache. The global cache setting limits the size of the cache, but you still need space for active processing of data segments (temp read/write) during transfers. I think that 10-15% is probably enough, but I've been recommending 20% to be safe.
I now have 10GB of free space on that drive which means about 5GB seems to have been permanently consumed beyond the cache itself. But I've run into another error "Client error: File does not exist." All VMs have failed and I'm creating a support ticket right now.
Thanks, Tom! I greatly appreciate the time you've spent answering my questions over the past few days.
I found the problem. I had an old folder sync job in place that was wiping out the WAN target folder. This folder sync was put in place to act as a backup copy job of sorts before I began using the WAN accelerator.
Would you consider adding the 15-20% overhead remark, made by Tom earlier, to the System Requirements for WAN Accelerators?
agreed - I am preparing to implement the WAN accelerators and was going to use a few 80GB SSD drives. Now that I see this I am thinking I need larger ones as I assumed I could consume the entire drive.
Could not have been a better timing for this request actually, as today I've been working on updating the main Release Notes document for the R2 release, fixing some omissions and making other general improvements along side... I was going to publish it tomorrow after final QC check, so there is still time for additions.
From what I know, I actually believe the value is going to be absolute - in other words, not dependent on the cache size. I will check with the devs to confirm.
Just wanted to note that the difference in WAN bandwidth savings between a 60GB and 80GB cache is unlikely to be more than a low single digit percentage, if that. In other words, if a 80GB cache gives you a 20x reduction, then the 60GB cache is likely to give 19.5x reduction, probably not worth worrying about. The difference between a 200GB cache and a 180GB cache is unlikely to even be measurable. A lot of users seem to think they need large global caches, but that's simply not the case unless you have a very high amount of significantly changing data, that still happens to somehow be repetitive, which just isn't any normal workload.
Gostev wrote:From what I know, I actually believe the value is going to be absolute - in other words, not dependent on the cache size. I will check with the devs to confirm.
I agree with this, I just wasn't 100% what the absolute value was. I'm thinking perhaps 10GB is safe but I wasn't sure if there were corner cases, so I'm interested to know.
We will be recommending additional 20GB of free disk space in cache folder per 1TB of source VM data.
To be scientifically correct, that is 3GB + 16GB per TB of source data.
Gostev wrote:We will be recommending additional 20GB of free disk space in cache folder per 1TB of source VM data.
To be scientifically correct, that is 3GB + 16GB per TB of source data.
This is over the top of conventional cache ? so if customer aims for 50GB cache and has 10TB of backups to move, they will need 250GB SSD for the cache folder?
This folder sync was put in place to act as a backup copy job of sorts before I began using the WAN accelerator.