NetEx shows performance data for HyperIP

NetEx has just released a report by DeepStorage Labs that provides some details about the performance improvements their HyperIP technology provides when migrating VMware virtual machines over WAN links. NetEx previously mentioned a sixfold improvement in VM migration times; according to the report the improvements were even better.

The main conclusions of the report are:

  • HyperIP accelerates VMotion across WAN links up to 10X
  • VMotion migration times over even low-cost networks are comparable to migrations within the data center
  • HyperIP’s virtual appliance architecture requires no changes to the underlying network simplifying implementation and reducing cost
  • HyperIP is a valuable addition to the system architect’s bag of tricks when data, or virtual systems, must be transferred between locations as in a DR scenario.

That’s very nice for the NetEx marketing department; but the full report provides some details about the technology used that I find far more interesting. Apparently the main problem NetEx is trying to solve is the fact that the TCP protocol doesn’t handle packet loss on high-bandwidth links in a very graceful manner (in most cases, the transmission speed is halved every time a packet is lost). The HyperIP software encapsulates the traffic in a UDP stream, using their own congestion control and a compression algorithm:

it sends the data across the WAN link between a pair of HyperIP appliances using UDP,
which is not latency sensitive.  NetEx adds a proprietary error correction protocol to UDP which
adjusts the amount of data “in flight” and block size to accommodate changing WAN conditions.
Rather than assuming that all packet loss is due to link congestion, HyperIP manages the
amount of data it sends avoiding congestion by adjusting the data sent rate.
In addition, HyperIP compresses data before segmenting it into packets to send across the net so
each packet can be sent at the optimal size. It also monitors the effectiveness of the compression
algorithm disabling compression when the compressed data is more than 80% the size of the source data.

[...] it sends the data across the WAN link between a pair of HyperIP appliances using UDP, which is not latency sensitive.  NetEx adds a proprietary error correction protocol to UDP which adjusts the amount of data “in flight” and block size to accommodate changing WAN conditions.

Rather than assuming that all packet loss is due to link congestion, HyperIP manages the amount of data it sends avoiding congestion by adjusting the data sent rate. In addition, HyperIP compresses data before segmenting it into packets to send across the net so each packet can be sent at the optimal size. It also monitors the effectiveness of the compression algorithm disabling compression when the compressed data is more than 80% the size of the source data.

Which would make the HyperIP software mainly useful when migrating VM’s over a link that has either high packet loss or low to medium packet loss combined with relatively high latency, which are exactly the scenario’s DeepStorage has tested.

Related posts:

  1. Lots of virtualization-related news
  2. Avere Systems shows first benchmarks
  3. SSD performance roundup by HotHardware

Leave a Reply

 

 

 

You can use these HTML tags

<a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <strike> <strong>