#RESILIO SYNC PIECES PRO#
#RESILIO SYNC PIECES FREE#

Configure R1Soft to perform a Block Level Backup of Server A to Server BSP and ensure to Verify the Backup. some data will not make the transfer if the site continues to be active during the server switch).

(3) Determine a window where you will find an 'acceptable loss of service' or 'acceptable loss of data' (i.e. R1Soft Agents will be deployed to Server A and Server B from Server BSP. Since I am more familiar with R1Soft that's what I will continue with. We have always used R1Soft but Acronis is another option. This server should be equipped with multiple NICs bonded and high speed storage. (2) A third server, we'll call it 'Server BSP' or Server Backup Storage Provider for easy reference.

As you are limited by the 1G switch, a LAG group with the second NIC having multiple GbE connections would be ideal. (1) Server A needs a second NIC of GbE or faster. What I am going to provide next is the best possible 'cheapest' way to go about this migration. Quite realistically: that's not going to happen - given the amount of drives required for a RAID6 that means there is at least a sizable enough storage to rightly assume. That is why I am looking for block level replication.Please correct me if I am wrong but: (1) you want a block level sync between the two servers for migration (2) the I/O average of server to be replicated is at 30%-40% sustained usually (3) there's one network card attached to it with a GbE link only (4) and you want to keep downtime to 1-2 hours based on the configuration you have provided. I don't mind running a sync job for days or even weeks, but the final downtime where the servers are completely synced, I wish to keep the downtime to less than 1~2 hours. When using rsync, I have tuned options(turn on/off compression, use fast ssh compression algorithms, limit bandwidth, limit IO - which of note doesn't work well for file system metadata), but it will take hours to rsync completely even after a initial rsync done beforehand. It is a RAID 6 consisting of common SATA drives with a cachecade RAID 1 SSD read/write acceleration caching. However server A has high load, with about 30~40% IO utilization and 400~500 Mbps traffic utilization.Īs the server is used as the main file system for one the top 500 worldwide sites by SimilarWeb traffic ranking, I wish to keep downtime minimal. Servers are currently on 1 GB internal network on same switch, although future plans to migrate to 10 GB switch.

#RESILIO SYNC PIECES FREE#

On a busy system your 1gigE port will likely already be saturated and your disks will be busy as well so that'll also reduce your rsync performance (random access is a lot slower on spinning disks).ĭoes your existing server have PCIe 2.0, 3.0 or 4.0 and do you have eSATA 6.0gbps, USB 3.x or free SAS ports? If you've got an older server with PCIe 2.0 that's going to affect performance. These are the maximum speeds, your real world performance will be less.

#RESILIO SYNC PIECES PRO#

The Samsung 970 PRO 1TB M.2 NVMe drive (PCIe 3.0) is spec'ed at 15,000 to 500,000 IOPS depending on the hardware configuration, up to 3,400 MB/s sequential reads and up to 2,500 MB/s sequential writes while using PCIe 3.0. SATA SSDs can provide about 10,000-85,000 IOPS depending on various conditions. Individual 3.5' spinning disks provide about 75-100 IOPS 7,200 RPM SATA, 125-150 IOPS RPM SATA, 120-160 IOPS RPM SAS, 175-220 IOPS RPM SAS, +- 10-20%Ģ.5' spinning drives are typically slower. You've got 2 bottlenecks, the first bottleneck is the RAID 6 with spinning disks and the second bottleneck is the 1gigE LAN.

Resilio sync pieces