WAN Acceleration technologies accelerate data transfers that eliminate data management challenges are known as ‘WAN acceleration.’ Bridgeworks patented artificial intelligence (AI) and machine learning (ML) technologies accelerate organisations existing technologies and dramatically accelerate their Wide Area Networks (WANs). AI underpins the patented technology to provide process intelligence, mitigating latency and mitigate packet loss. Our solutions, such as PORTrockIT, dramatically improve data throughput up to 98% of bandwidth – regardless of distance.
Accelerating BlackPearl, Securing Data
Spectra Logic: “In our tests, PORTrockIT accelerated BlackPearl’s replication performance significantly – in one case, completing the job more than 5 times faster.”
BlackPearl is a Spectra Logic platform that aims to make it easy for organisations to preserve data simply, affordably, and safely. It provides a RESTful interface that allows users to store files as objects and uses an onboard cache to accelerate the movement of data on and off long-term storage media, such as disk arrays and tape libraries.
One of the key use cases for BlackPearl is to enable geographically dispersed archiving, where data is preserved at multiple sites to reduce the risk of data loss in the event of a disaster. BlackPearl makes this easy to achieve by simply connecting two BlackPearl devices across a wide area network (WAN) connection.
However, whenever data is transferred across a WAN, it is subject to the effects of latency and packet loss, which can have a dramatic impact on data throughput. Using BlackPearl together with PORTrockIT can transform the performance of data replication across the WAN by counteracting the impact of both latency and packet loss. In every scenario we evaluated, PORTrockIT was able to increase transfer rates significantly. In one case, it reduced the total transfer time from 247 minutes to just 47 minutes, completing the job more than five times faster.
A complete complement
BlackPearl can help businesses to create and maintain geographically dispersed archives, while PORTrockIT adds value by mitigating the risk of losing valuable data and making it easier to demonstrate compliance with data retention regulations.
To demonstrate the results that PORTrockIT can deliver for BlackPearl customers, Bridgeworks conducted a set of performance tests at a testing facility in the UK. The test infrastructures mimicked a real-world BlackPearl architecture, using an “Injector” server to inject or extract files from BlackPearl, and a WANulator to simulate different levels of latency across a wide area network. Performance figures were taken from the BlackPearl console and recorded.
The tests assessed WAN data transfer performance at various simulated levels of latency. The WANulator also simulated a nominal packet loss rate of 0.1%, representing a typical service-level agreement from most WAN providers.
Test 1 and Test 2
Test 1 and Test 2 were run first with PORTrockIT set to bypass mode, to simulate an unaccelerated network architecture. The tests were then rerun with PORTrockIT in active mode, to simulate an accelerated network architecture. The figures from the two test runs were then compared to assess the difference in performance between the unaccelerated and accelerated architectures.
Test 3 - a slightly different setup
Test 3 used a slightly different setup, with a pair of Bridgeworks WANrockIT appliances in place of the PORTrockIT appliances. This is because WANrockIT is specifically designed to enable the transfer of SAN block-level protocols across a WAN, which is necessary for testing the specific use case of transferring data between a local BlackPearl and a remote tape library that are linked by a fibre channel connection.
Across all three tests, the most important metrics to consider are:
•Average transfer rate, measured in megabytes per second. This allows us to assess the speed at which files can be transferred across the WAN.
•Total time to completion, measured in minutes. This allows us to judge how long the entire file transfer job takes.
Test 1 assessed the performance of transferring a 1 TB payload of files from the local BlackPearl cache to the remote BlackPearl cache and back again at 20, 40, 60, 80 and 100 ms of latency with 0.1% packet loss.
Test 2 extended Test 1 by adding a tape library connected to the BlackPearl appliance at the remote end of the WAN. This simulated a scenario where BlackPearl is used as a distributed archive for long-term data retention.
In Test 3, one BlackPearl appliance was placed on the local side of the WANulator and connected via a fibre channel connection to a remote tape library, using WANrockIT as the accelerator. This would evaluate the speed of reading directly from tape across the WAN.
Accelerated Results
Without PORTrockIT - For the unaccelerated network architecture, the results of Test 1 and Test 2 show a negative correlation between latency and performance. As latency increases, average transfer rates (ATR) decline and we see a longer total time to completion (TTC). For example, in Test 1, the TTC was 119 minutes when there was 20 ms of latency but rose to 246 minutes at 100 ms of latency – more than twice as long.
The results of Test 2 were similar: the TTC was 78 minutes at 20 ms of latency, but 157 minutes at 100 ms of latency – again, more than double the time to complete the transfer. In a real-world scenario, this variability in TTC would be a major concern, as any unexpected increase in latency could cause transfers to time-out or overrun their scheduled window – putting valuable business data at risk.
With PORTrockIT - When we look at the architecture accelerated by PORTrockIT, the results are very different. At all levels of latency, the ATR improves dramatically when PORTrockIT is activated. For example, in Test 1, the worst-case ATR of the accelerated architecture (365 MB/s at 100 ms of latency) is still more than 2.5 times as fast as the best-case ATR of the unaccelerated architecture (145 MB/s at 20 ms of latency). We can also see that PORTrockIT keeps TTC stable, regardless of latency.
In Test 1, the shortest TTC was 39 minutes at 20 ms of latency, and the longest was 46 minutes at 100 ms of latency. In Test 2, the fastest TTC was 36 minutes and the slowest 43 minutes. In both cases, that’s a difference of just 7 minutes between the slowest and fastest transfers – which suggests that real-world users can expect their transfers to complete on time, even if latency increases.
Test 3 demonstrates that Bridgeworks WANrockIT, a WAN acceleration product for SAN protocols, Spectra Logic customers cannot only geographically separate their BlackPearl instances, but also use a third geographical site for air-gapped tape storage facilities without suffering the performance penalties associated with latency and packet loss.
Accelerating IT security
The project with the major IT security firm is also a game-changer. Between November and December 2021, a storage subject matter expert within this company came to Bridgeworks with details about a file transfer problem on a 2Gbps WAN circuit between Colorado, USA, and Bangalore in India.
Problem:
With data equating to 700TB storage estate, backups are only as good as the ability to restore the data. In order to allow an acceptable transfer time of this data requires a robust solution.
There were 2 NetApp units involved in the data migration: one in Bangalore and another in Denver. The transfer was performed using the inbuilt NetApp Snap Mirror feature. It involved migrating data in a follow the sun requirement, and critical data and security was paramount. The allocated WAN bandwidth for transfer was a 1Gb/s controlled by a SD-WAN. But the raw WAN bandwidth was 2 Gb/s. Before PORTrockIT, throughput was typically equating to 2MB/s using Incumbent technology WAN-Optimisation with only 2% unitisation of WAN, and the time to transfer the data was 8.4 days.
After installing and configuring the Bridgeworks PORTrockIT virtual instances, the transfer speed increased to 80MB/s – a 40-fold increase and the time taken was reduced to under 6 hours. In this case the benchmark figure we were given was that the customer was only able to move data at around 2 Mega Bytes per second.”
Results
A spokesperson from the IT security company remarks: “The rate of throughput was incredible. PORTrockIT from Bridgeworks is a game changer. I am a storage engineer with limited networking knowledge. I was able to configure and deploy PORTrockIT in less than 2 hours”. The IT security firm believes that with ‘unthrottled’ conditions and utilising their full 2 Gig WAN, they would be able to transfer data at 200 MBs per second.
PORTrockIT is working over a SD-WAN, and they required no other changes to the LAN or WAN configurations.
Speaking about the project, Kevin St. Cyr, Tributary Systems inc. of Dallas, Texas says: “We have always impressed with the capabilities of PORTrockIT product, but what they have achieved with the IT security company’s installation takes it to another level. It is a showcase for the capabilities of PORTrockIT. The beauty is, it’s easy to install and it shows the performance that can be achieved over a WAN that was once thought impossible.”
WAN Acceleration: Accelerating cloud metro cluster deployment
WAN Acceleration technologies accelerate data transfers that eliminate data management challenges are known as ‘WAN acceleration’. Bridgeworks patented artificial intelligence (AI) and machine learning (ML) technologies accelerate organisations existing technologies and dramatically accelerate their Wide Area Networks (WANs). AI underpins the patented technology to provide process intelligence, mitigating latency and mitigate packet loss. Our solutions, such as PORTrockIT, dramatically improve data throughput up to 98% of bandwidth – regardless of distance.
Hybrid cloud, object storage and metro clusters
Steven Umbehocker (SU), CEO and CTO of OSNEXUS Corporation explains why this is important to organisations that are accelerating their adoption of hybrid cloud strategies: “This can be in the form of object storage for backup, disaster recovery sites, and all this puts a greater importance on efficient use of the WAN. This is important across all sectors. However, it is especially important in sectors such as healthcare to protect data like PACS images.”
Before implementing Bridgeworks PORTrockIT OSNexus was facing performance challenges in deploying metro clusters and disaster recovery (DR) sites over WANs – suffering with particularly high latency over large distances. There was a 200-700% boost in performance by implementing PORTrockIT which is a game changer. The highest performance improvement was 7 times higher than without any WAN acceleration.”
Critical: Recovery Time objectives
Recovery time objectives (RTOs) are critical to IT organisations, so it is imperative to have adequate WAN bandwidth and throughput to keep the remote replication to DR sites within the required RTO. OSNexus is also seeing increase demand for metro clusters where the storage is distributed across multiple sites to achieve zero downtime in the event of a site outage.
Umbehocker therefore adds: “That’s only possible when the latency is low enough to provide sufficient and adequate performance for the workloads. PORTrockIT solves these latency issues so that metro clusters and DR sites can be deployed across a much larger geographic area. If you have an earthquake in one zone, that larger metro cluster offers benefits. You can avoid downtime, enabling organisations to link their disparate sites together. There is much latency between New York and Tokyo, and so you need something like PORTrockIT to mitigate latency.”
With high latency and packet loss, he explains that data transfers are delayed and may need to be resent which reduces performance – the transfer rate. WANs are typically spanning portions of the public internet and/or shared dark fibre infrastructure that IT organisations don’t have much control over, and simply upping bandwidth doesn’t address the latency issues. What is required is a different approach to boosting WAN performance that is designed to specifically address the inherent latency issues and packet loss.
Before approaching Bridgeworks, OSNexus experimented with various network tuning options, including adjusting the frame sizes and queueing techniques, but these adjustments do not come anywhere close to the range of benefits offered by PORTrockIT. The three biggest benefits were the reliability of the link, the visibility of the performance, and increased performance with PORTrockIT.
Joshua Blake (JB), Regional Support Lead – EMEA at OSNexus Corporation therefore says: “It definitely increased performance, but I couldn’t tell you if there was more smoothness. It was easy to set up once you understand it. There were some things I needed to read up on, but it was otherwise easy to set up.”
The company has a disaster recovery site between Seattle and Las Vegas, and so it found it helpful to include PORTrockIT into their automated test systems. To that extent, PORTrockIT is now certified for use with QuantaStor, and this enables OSNexus customers to leverage it for disaster recovery sites and metro cluster use cases.
Performance improvements
The QuantaStor VM endpoints shown in Figure 1 use the PORTrockIT virtual appliances as gateways for replicating data from a ZFS file system. For the data payload, OSNEXUS used a randomly generated 5,124 MB test file. The ZFS data was sent and received via SSH on TCP port 22.
By using WANulator to simulate the scenarios presented in Table 1, the OSNEXUS team was able to test transfer speeds both with and without PORTrockIT data acceleration and make direct comparisons between the two data transfer strategies. The test was conducted five times for each different latency, and the average transfer time in seconds can be seen below. Packet loss equated to 0.1%, and latency ranged from 0 milliseconds to 160 milliseconds.
At 0ms the unaccelerated the unaccelerated transfer time equated to 107.0 seconds. When accelerated this amounted to 105.6 seconds, with an improvement factor of -0.01. At 20ms the unaccelerated transfer time was 333.6 seconds, while the accelerated time was 117.6 seconds – leading to an improvement factor of 2.83. At 40ms the unaccelerated transfer rate was 1,724.6 seconds, and the accelerated transfer time was 245.0 – an improvement factor of 7.03. At 80ms the unaccelerated transfer time sat at 2,945.4 seconds, while accelerated it stood at 454.6 seconds – an improvement factor of 6.47. Lastly, at 160ms the unaccelerated transfer time was 3,789.0 seconds. Yet when accelerated it stood at 871.6 – an improvement factor of 4.34.
Blake therefore comments: “Aside from the performance benefits, being able to identify as a complementary product – architecting it with storage – adds another string to our bow effectively. We have mentioned PORTrockIT as a goal. We go in to speak about it, and then they see the benefits themselves – it’s like planting a seed.”
OSNexus has clients that are implementing hybrid cloud strategies who want or need the benefits that, according to Umbehocker, need to the benefits that PORTrockIT solves. This is because PORTrockIT allow data to be moved over WANs quickly and efficiently is a core challenge in implementing hybrid cloud strategies.
He concludes: “We have already started adding more features like automatic tiering from file to object storage and these new features around replication and tiering of data are putting an increased importance efficient use of customers available WAN bandwidth. PORTrockIT enables organisations to get increased value and performance out of their monthly WAN bandwidth investment, with an increase in throughput up to a 700%.”
National Institutes of Health
The NIH was originally a PORTrockIT 400 customer. However, it found an additional need to add two of Bridgeworks’ new high end PORTrockIT 800’s last year. They are moving four protocols over the WAN at ‘breakneck speed’.
A spokesperson from BKJ Works - New York comments: “This is a non-engineering group, but they have had no problem in installing the devices (each in under 1 hour!). Don't be surprised if they come up with future needs on an even grander scale in the future. I cannot thank Bridgeworks enough for their continued support, fast turn around and meeting all and any of our requests with a willingness to say, ‘Yes we can’. This relationship will only deepen going forward.”
WAN Data Acceleration powers the National Institutes of Health’s global data transfer requirements at the U.S Department of Health and Human Services. It is also the world’s largest biomedical research agency. One Bridgeworks' healthcare customer, US-based CVS Caremark, connected two virtual tape libraries over 2,860 miles at full WAN bandwidth. This achieved a performance gain of 95 times the unaccelerated performance.
The National Institutes of Health (NIH) has gone from never being able to achieve fast datacentre replication over a distance between sites of 2,000 miles, to be able to attain it. The difference between doing it between two machines in the same datacentre and in two machines across two datacentres was 6 MB/s on NetApp Snap Mirror replication. So, they now lose only this amount of time.
The NIH wasn’t able to address this issue for a couple of years. NetApp and Aspera couldn’t achieve this. Bridgeworks was brought in to conduct a quick POC that took a couple of days to set it up, and then that was it. The savings come from not having anything previously to protect their data. Bridgeworks has also helped NIH to resolve network infrastructure issues. Although successful, the results aren’t yet in the public domain. Yet it is a substantial multi-million U.S. dollar contract.
Accelerating IT security
The project with the major IT security firm is also a game-changer. Between November and December 2021, a storage subject matter expert within this company came to Bridgeworks with details about a file transfer problem on a 2Gbps WAN circuit between Colorado, USA, and Bangalore in India.
Problem:
With data equating to 700TB storage estate, backups are only as good as the ability to restore the data. In order to allow an acceptable transfer time of this data requires a robust solution,
There were 2 NetApp units involved in the data migration: one in Bangalore and another in Denver. The transfer was performed using the inbuilt NetApp Snap Mirror feature. It involved migrating data in a follow the sun requirement, and critical data and security was paramount. The allocated WAN bandwidth for transfer was a 1Gb/s controlled by a SD-WAN. But the raw WAN bandwidth was 2 Gb/s. Before PORTrockIT, throughput was typically equating to 2MB/s using Incumbent technology WAN-Optimisation with only 2% unitisation of WAN, and the time to transfer the data was 8.4 days.
Phil Hill, Customer Solution Architect at Bridgeworks explains: “Our PORTrockIT solution for this company mitigated the effects of Latency and Packet loss. He says the latency between Denver and Bangalore at 350ms is a lot of latency! The biggest single factor causing latency is distance - even though the speed of light is around 186,282 miles per second – this figure is actually around one third lower when going through fibre optic cables.
Although the units can be hardware, sometimes virtual images or cloud machine images are more useful. As time was of the essence and the customer had real looming deadlines, the IT security firm opted for a Virtual Machine (VM) running in VMWare ESX. This made the turnaround for the evaluation-solution really quick. The following table shows the technical specification of the virtual products which would server the customer best.
After installing and configuring the PORTrockIT virtual instances the transfer speed increased to 80MB/s – a 40-fold increase and the time taken was reduced to under 6 hours. In this case the benchmark figure we were given was that the customer was only able to move data at around 2 Mega Bytes per second.”
Results
A spokesperson from the IT security company remarks: “The rate of throughput was incredible. PORTrockIT from Bridgeworks is a game changer. I am a storage engineer with limited networking knowledge. I was able to configure and deploy PORTrockIT in less than 2 hours”. The IT security firm believes that with ‘unthrottled’ conditions and utilising their full 2 Gig WAN, they would be able to transfer data at 200 MBs per second. PORTrockIT is working over a SD-WAN, and they required no other changes to the LAN or WAN configurations.
Speaking about the project, Kevin St. Cyr, Tributary Systems inc. of Dallas, Texas says: “We have always impressed with the capabilities of PORTrockIT product, but what they have achieved with the IT security company’s installation takes it to another level. It is a showcase for the capabilities of PORTrockIT. The beauty is, it’s easy to install and it shows the performance that can be achieved over a WAN that was once thought impossible.”
WAN Accelerating F1
A Formula 1 (F1) team undertook an evaluation, a proof of concept (POC), which showed the ability of PORTrockIT to make their WAN data transfers 60 times faster than before. The trial was so successful that the team ended it early. Bridgeworks’ brief was to help them speed up data the transfer of data from the Cotswolds in the UK to New England, in the USA.
A new Partnership with Prodec Networks was formed in the process, and a spokesperson from Prodec comments: “Regarding the F1 team, from wind tunnel analysis to engineering documents, Formula 1 teams accumulate significant volumes of data. The cars’ sensors (over a hundred of them) tell their own story about the race for the duration of the race and for referral later.”
“Needless to say, up to date revisions of the cars’ specific designs need to be easily accessible. How the teams deal with the pressures of reacting quickly to the vast amount of information is often the difference between winning and losing each race. With race data exceeding 150Gb per race weekend, it’s easy to see how our F1 team needed a fast data-transfer solution.”
Bridgeworks continues to keep its customers on track by enabling them to mitigate latency and packet loss. Even organisations that are implementing SD-WANs can benefit from a WAN Acceleration, which is data agnostic, and which permits the transmission of encrypted data at speed over large distances. This is something that tools such as WAN Optimisation will never be able to achieve.
