Feature Bridgeworks has found way to improve wide area TCP/IP transmission of large, streamed files that compensates for packet loss, radically improving the performance of NetApp SnapMirror and SnapVault which are hobbled by packet loss.
It’s been working on the technology to virtualise a TCP/IP connection for several years.
Here’s what we said in 2009: “You can have more than one logical TCP/IP connection on a link ... Its idea is to send a packet across a link, then open another connection and send a packet on that, and to keep opening connections until you get an acknowledgment (ack) on the first connection and send the next packet in the sequence. That way, you increase the number of packets in transit on the wire.”
With products such as NetApp’s SnapMirror, data is replicated across a network from one FAS array to a distant one, for protection purposes. If the network is noisy then packets can be lost, and minimal packet loss can severely impact performance.
Bridgeworks, which calls itself a "leaders in software defined protocol acceleration ... moving data irrespective of size, protocol or distance at high speed over the planet’s TCP/IP infrastructures", says packet loss can be a source-to-target data packet or a target-to-source acknowledgement (ack).
TCP/IP then automatically reduces the number of packets it sends in the next group, to compensate for the unreliability of the connection. So network utilisation is reduced, because the sender is sending fewer packets in the same amount of time.
Traditional WAN optimisation involves reducing TCP/IP chatter, deduplication and/or compression, all of which reduces the amount of traffic sent across the network, and none of which will do anything about packet loss, or so we're informed by David Trossell, the CEO of Bridgeworks.
His PORTrockIT product virtualises a physical TCP/IP wire into several, concurrent software connections. It coalesces an incoming stream of packets into chunks and sends them across software connections in parallel. Packet loss affects the SW connection in which it occurs and TCP/IP deals with it whilst other software connections continue operating.
A piece of patented Bridgeworks AI code in PORTrockIT analyses the state of the network and optimises chunk size and the number of software connections to maximise network throughput. For example, if the packet loss rate increases it will increase the number of software connections, reducing the number if the packet loss rate decreases.
Bridgeworks claims the combination of PORTrockIT TCP/IP wire virtualisation and the AI-based optimisation produces a Self-Configuring, Infrastructure-Optimised Network (SCION) product. This can, Bridgeworks adds, accelerate the data to 95-98 per cent network utilisation. Throughput is higher than before, performance is more constant, and file transfer time more predictable.
PORTrockIT runs in an x86 appliance or as a virtual machine inside an x86 server, and it sits between the data sending source at one end of the link and in front of the target at the other end of the link. It works whether the incoming data stream is encrypted or not as it does not look into the packet contents.
This software comes into its own when a noisy network, one that causes packet loss, is being used. With even just 0.5 per cent pack loss with a large file transfer the cumulative effect is disastrous in terms of file transfer time. By effectively sub-dividing the large file into smaller ones sent in parallel across the same physical link, packet loss effects are limited in scope and overall file transfer time across the affected network increased dramatically.
Bridgeworks has charted what happens to the WAN's transfer rate when packets are lost. We show the effect at zero, 0.5 and 1.5 per cent packet loss, with NetApp SnapMirror running un-aided and accelerated with PORTrockIT;
Accelerated and unaccelerated performance at various latencies with zero packet loss; effectively no difference.
Accelerated and unaccelerated performance at various latencies with 0.5 per cent packet loss. PORTrockIT accelerates transfer by 6-50X or more, with acceleration increasing with latency.
Accelerated and unaccelerated performance at various latencies with 1.5 per cent packet loss. This chart shows a 60X or more improvement at thelongest latency.
Bridgeworks says in the mist extreme example, 150msecs of latency with 1.5 per cent packet loss, PORTrockIT acceleratioin achieved a 68MB/sec transfer rate compared to less than 1MB/sec with the unaccelerated TCP/IP link.
If you have such a lossy network affecting your WAN data transfers them imagine what an up to 68 X decrease in transfer time for large files might mean.
Bridgeworks' technology can enable replication frequency to be increased, thus improving your recovery point objectives.
PORTrockIT is available today and works with FTP, REST, NetApp SnapMirror and SnapVault, Veritas NetBackup, IBM Spectrum Protect, Caringo SWARM and Datacore. Contact Bridgeworks for more information. ®