Articles

After 2 weeks of data from our Steelhead 2050 install, I'm super impressed with what the steelheads are capable of...but a warning, make sure you size the device properly for the amount of data you're going to be moving. The steelheads we just installed are reducing data over the WAN link by 95%, but now we're running into a disk I/O bottleneck. The steelhead disks cannot keep up with the SAN on the LAN side.

• SATA disk performance is bad, especially with only 4 SATA spindles. Using default SDR and adaptive SDR mode we see very high disk pressure, data reduction over the WAN is awesome, 95% reduction for equallogic SAN replication. But the stealhead disks cannot keep up with the SAN, over the course of a week replication times on the equallogic degraded and disk pressure on the steelheads went up. Turning on SDR-M cut replication times in half, but obviously hurt data reduction since we're only using 6GB of RAM instead of 400GB of disk.
  
  for $40,000 units, why is SATA used?
 
• Right now the model sizing is done by number of TCP connections, WAN throughput, and storage capacity. I don't need more TCP connections, or WAN throughput or storage capacity, I need faster disk.  What's the point of spending another $40,000 to jump up a model that only gives me two more SATA disks and a bunch of stuff I don't need?


• Why not build units just dedicated for site to site data replication?
  • Fast disk/the ability to use your own DAS
  • No RSP BS (seriously, who the hell is going to run a VM on one of these?)
  • More memory
  • Less TCP connections