High Speed Mass Encryption/Decryption Testing
Using Kryptera Mirage & Enterprise

1. Summary

We worked with the Centre of Excellence in Next Generation Networks (CENGN) to test the stability and speed of Mirage server-side and client-server file and directory processing. We also performed a subset of server-side tests using the Kryptera Enterprise server.

We performed our server-side testing on a bare metal Cisco UCS M5 server that was provided, configured and hosted by CENGN. The server was equipped with 384GB of memory, 800 GB of NVMe storage, two Intel Xeon Gold 6140 processors with 36 cores/72 threads, and a 10 Gbit Network Interface Card (NIC).

Our goal for Mirage was to create baseline processing speeds, with one, two and three client servers that transferred data to and from the Mirage server. We created custom software with custom server configurations to perform the client side tests to reach maximum transfer speed across the network. We did this because scp and sftp are limited to using one core at a time, where time-slicing is used to poorly mimic multi-threaded processing. Our software used all cores to transfer files to and from the Mirage server across the 10 Gbit network.

At the time of testing, Mirage and KES used a single 4,096 bit symmetric key. The key size was determined by the architecture of the product line, and has been reduced in size since that time to 384 bits. A single additional key bit doubles the complexity and time needed to break a private key. We reduced the key size without any risk to security. The private key is subsequently mutated and then expanded in size before use as a pool to be drawn from as needed.

Mirage provided 8-bit modulus encryption. After key mutation and expansion, a fixed number of substitution tables were randomly selected from a maximum range of over 500 million highly unique substitution tables. Each byte within each file was substituted from one randomly drawn substitution tables from the randomly chosen tables, where table use was randomly determined. After substitution the resulting byte was randomly permutated before it was written to the target file. The random permutation values are randomly changed after use, as are the random sequencing of selected substitution tables. Since refactoring our code base, the expanded key pool is used to draw random bytes as needed, with subsequent key mutation and expansion to refresh the pool on depletion.

KES was configured to provide 16-bit modulus encryption, which allows access to a maximum range of over 2 billion highly unique substitution tables. Each file object is substituted from a randomly determined substitution table. After substitution the resulting object is randomly permutated then written to the current block. The completed block is randomly scrambled using a derivative of the symmetric key before being written to the target file. Each block uses a randomly altered expanded symmetric key.

Mirage uses a different approach than KES to encryption, and was designed to achieve maximum speed. KES and Mirage both have a 63-bit upper limit on file size. KES uses high-speed encryption and decryption equations, where Mirage uses algorithms to create substitution tables. Both use different approaches to utilizing substitution tables, and Mirage front-loads all work to speed up highly-dedicated file processing.

Our fastest server-side processing time for Kryptera Mirage was 1.63 billion bytes/second when decrypting 12 25 GB files.

Our fastest server-side processing time for the Kryptera Enterprise Server was 885.8 million bytes/second when encrypting 2,050 150 MB files.

The fastest client-server processing time Kryptera Mirage was 482.5 million bytes/second when encrypting 1,000 100 MB files.

We also tested the processing speed of AES encryption using OpenSSL on the server. It reached a maximum speed of 419 million bytes/second when encrypting a single 5 GB file. OpenSSL relies on AES-NI to encrypt and decrypt files, where AES-NI is fully integrated into Intel and other manufacturers’ processors. It operates at the same speed as the processors.

During testing of Mirage we realized that processing speed is fully at the mercy of core heat, and the way that the processors and server reduce heat by reducing core frequency and throttling file I/O. While all fans were set to their maximum value to reduce heat as much as possible, this proved insufficient. Our server side speed tests should have proven predictable because the processing time of encryption and decryption of the same sized file should prove identical. We monitored core use and core frequency during extended levels of file processing, which proved that cores were often operating at low levels of utilization purely to reduce heat.

In future, we will deploy liquid cooled Mirage servers to further reduce heat and maximize the processing speed of Mirage. We will also determine if processing speed can be greatly improved through use of GPUs.

We found that the Kryptera Enterprise Server was operating at maximum core speed for the duration of file processing apparently without being held back by the processors by reducing core frequency and throttling I/O to reduce heat. This indicates that KES will not be sped up to any great degree by using liquid cooled servers. The only way to increase the speed of KES will be to use faster, and a greater number of, CPU or GPU cores. Mirage can be sped up by reducing core temperature as much as possible, and by using faster, and a greater number of, CPU and/or GPU cores

2. Mirage, KES & AES-NI Server-Side Speed Tests

Hyper-threading was disabled in the BIOS to permit each core to host a dedicated thread to improve performance. The Intel 6140 processors have a base frequency of 2.30 GHz, a minimum frequency of 1.00 GHz, and a maximum turbo frequency of 3.70 GHz. We encountered heat issues when threading out processing across all cores. The fans were set to their maximum speed of 12,000 RPM to reduce heat as much as possible. A few other BIOS refinements were made to improve processing speed. We ran into media errors on the NVMe storage. To get around this problem, the partition was deleted, a new partition was created, and the media was repeatedly low-level formatted until the errors were correctly mapped.

The processing speed of encryption and decryption of a specific file should always require the same amount of processing time. Heat played a major role in reducing the speed of processing for all of our tests.

In the following test results, the base line processing speed that proves highest for a specific number and size of files should be considered as accurate even though the speed was reduced due to the processors reducing core speed, and throttling file I/O speed, to lower heat and use of electricity.

In the following table we also present a subset of speed tests using the Kryptera Enterprise server, and the speed of performing AES encryption using OpenSSL on individual files that ranged in size from 300 GB to 750 MB. These tests were performed to determine how fast Mirage could process multiple files compared to AES-NI processing a single file at a time. AES-NI is fully integrated into Intel and other manufacturers’ processors, and operates at the same speed as the processors. We encrypted each file using OpenSSL, which passed the request to AES-NI for processing. OpenSSL is designed to encrypt or decrypt a single file at a time, like most open source and commercial encryption products.

Note that we used 35 of 36 available cores to allow the primary thread to operate at the highest possible speed. We also used a delay period of 1/1000th of a second between files queued up for processing, with compensation to prevent improper reuse of thread work space. The speed of Mirage file processing is such that volatile memory used to track thread activity occasionally lags under heavy load. We incorporated a second test to ensure that a working thread was not used incorrectly. This change ensured that we could lower the delay period to 1/1000th of a second without any risk of dropping a file due to improper thread use.

We have provided the maximum processing speed within the following table because it is most accurate. The speed of Mirage encryption and decryption should be identical when comparing a specific number of files and size of files. The actual differences in recorded processing speed are due to the way the server was trying to control heat and use of electricity by lowering core frequencies, swapping cores, and throttling the speed of file I/O.

Files Size Maximum Processing Speed Notes
1 300 GB 200 million bytes/sec Mirage
1 300 GB 353 million bytes/sec AES-NI
1 200 GB 220 million bytes/sec Mirage
2 150 GB 443 million bytes/sec Mirage
1 150 GB 226 million bytes/sec Mirage
1 150 GB 404 million bytes/sec AES-NI
3 100 GB 623 million bytes/sec Mirage
1 100 GB 228 million bytes/sec Mirage
1 100 GB 401 million bytes/sec AES-NI
4 75 GB 685 million bytes/sec Mirage
1 75 GB 231 million bytes/sec Mirage
1 75 GB 411 million bytes/sec AES-NI
6 50 GB 910 million bytes/sec Mirage
1 50 GB 230 million bytes/sec Mirage
1 50 GB 410 million bytes/sec AES-NI
10 30 GB 1.26 billion bytes/sec Mirage
1 30 GB 232 million bytes/sec Mirage
1 30 GB 416 million bytes/sec AES-NI
12 25 GB 1.63 billion bytes/sec Mirage
1 25 GB 229 million bytes/sec Mirage
1 25 GB 413 million bytes/sec AES-NI
15 20 GB 1.61 billion bytes/sec Mirage
1 20 GB 231 million bytes/sec Mirage
1 20 GB 418 million bytes/sec AES-NI
20 15 GB 1.50 billion bytes/sec Mirage
1 15 GB 230 million bytes/sec Mirage
1 15 GB 413 million bytes/sec AES-NI
30 10 GB 1.53 billion bytes/sec Mirage
1 10 GB 233 million bytes/sec Mirage
1 10 GB 413 million bytes/sec AES-NI
60 5 GB 1.46 billion bytes/sec Mirage
60 5 GB 763.3 million bytes/sec KES
1 5 GB 224 million bytes/sec Mirage
1 5 GB 419 million bytes/sec AES-NI
100 3 GB 1.43 billion bytes/sec Mirage
100 3 GB 865.7 million bytes/sec KES
1 3 GB 230 million bytes/sec Mirage
1 3 GB 403 million bytes/sec AES-NI
150 2 GB 1.52 billion bytes/sec Mirage
150 2 GB 789.5 million bytes/sec KES
1 2 GB 219 million bytes/sec Mirage
1 2 GB 413 million bytes/sec AES-NI
300 1 GB 1.44 billion bytes/sec Mirage
300 1 GB 865.0 million bytes/sec KES
1 1 GB 195 million bytes/sec Mirage
1 1 GB 30.7 million bytes/sec KES
1 1 GB 398 million bytes/sec AES-NI
410 750 MB 1.38 billion bytes/sec Mirage
410 750 MB 873.8 million bytes/sec KES
1 750 MB 197 million bytes/sec Mirage
1 750 MB 31.5 million bytes/sec KES
1 750 MB 393 million bytes/sec AES-NI
614 500 MB 1.31 billion bytes/sec Mirage
614 500 MB 877.1 million bytes/sec KES
1,024 300 MB 1.18 billion bytes/sec Mirage
1,024 300 MB 870.6 million bytes/sec KES
1,230 250 MB 1.11 billion bytes/sec Mirage
1,230 250 MB 869.1 million bytes/sec KES
1,538 200 MB 1.25 billion bytes/sec Mirage
1,538 200 MB 878.9 million bytes/sec KES
2,050 150 MB 1.38 billion bytes/sec Mirage
2,050 150 MB 885.8 million bytes/sec KES
3,075 100 MB 1.27 billion bytes/sec Mirage
3,075 100 MB 866.8 million bytes/sec KES
4,100 75 MB 1.26 billion bytes/sec Mirage
4,100 75 MB 869.1 million bytes/sec KES
6,150 50 MB 1.16 billion bytes/sec Mirage
6,150 50 MB 841.9 million bytes/sec KES
10,250 30 MB 1.20 billion bytes/sec Mirage
10,250 30 MB 816.3 million bytes/sec KES
12,300 25 MB 1.16 billion bytes/sec Mirage
12,300 25 MB 804.1 million bytes/sec KES
15,375 20 MB 1.13 billion bytes/sec Mirage
15,375 20 MB 784.5 million bytes/sec KES
20,500 15 MB 1.28 billion bytes/sec Mirage
20,500 15 MB 749.9 million bytes/sec KES
30,750 10 MB 1.00 billion bytes/sec Mirage
30,750 10 MB 688.9 million bytes/sec KES

3. Mirage Client-Server Speed Tests

Client side testing was performed on a virtualized Cisco UCS M5 server. Each virtualized Ubuntu server was configured with 8 cores using Intel Xeon Gold 6140 processors. A 10 Gbit NIC was set up to directly link to the Mirage server. Two of the VMs were configured with 330 GB of NVMe storage, where the third was configured with 205 GB of SSD storage. The file write speed of the SSD storage was 60% of the write speed of the NVMe storage due to the bus limitations. This speed difference caused testing using all three VMs at the same time to quickly become out of sync. At times, client3 processed files at a faster rate than client1 and client2 because of random events such as delays caused by waiting for Mirage file processing to complete thus allowing client3 to fully utilize the network.

We developed software to automate file creation, timed transfers of files to and from Mirage, and discretionary verification of source files with associated decrypted files. We performed file verification for several hundred files without difficulty, and disabled further testing to reduce the time needed for each series of tests.

The first timed test transferred original files to Mirage to encrypt. Once all files were transferred to Mirage for processing, all cores were repeatedly used to wait until specific files had been encrypted before transferring them back to the client VM. While Mirage processes files as quickly as possible, the speed tests for encryption entail transferring all files across the network to Mirage, then waiting for each file until it could be transferred back to the client VM. This dual transfer plus wait greatly reduces the apparent speed of file processing compared to the speed of server side files processing.

The second timed test transferred encrypted files to Mirage to decrypt. Once all files were transferred to Mirage for processing, all cores were repeatedly used to wait until specific files had been decrypted before transferring them back to the client VM.

During testing we discovered that OpenSSL, which is the core service used for ssh, sftp and scp, does not permit multi-threading of data transfers. Instead, it performs time-slicing, which gives each core a percentage of total time to handle data transfers. This obsolete functionality greatly reduces the maximum speed of file transfers, and absolutely negates the value of a mass encryption solution such as Mirage.

We worked around this problem by mounting specific directories shared by the Mirage server on each client VM. Once mounted, we enhanced our client test software to fully utilize available cores to thread out transfers of files to and from the Mirage server. We also made a change to the Mirage daemon to use semaphores to allow our client-side software granular control over when file processing would occur. This change to using a semaphore system rather than rely on the Mirage daemon checking for associated file handles greatly sped up processing time for smaller files. We retested the server side tests after the client tests were complete and realized that processing time for files that range in size from 10 MB to 200 MB were greatly improved.

Files Size Encryption Decryption Notes
1 100 GB 75.6 million bytes/sec 74.8 million bytes/sec client1
1 100 GB 76.0 million bytes/sec
66.9 million bytes/sec
73.0 million bytes/sec
63.3 million bytes/sec
client1
client2
1 100 GB 70.5 million bytes/sec
62.2 million bytes/sec
52.0 million bytes/sec
65.2 million bytes/sec
45.8 million bytes/sec
53.7 million bytes/sec
client1
client2
client3
2 50 GB 132.0 million bytes/sec 126.9 million bytes/sec client1
2 50 GB 118.4 million bytes/sec
109.9 million bytes/sec
117.0 million bytes/sec
110.7 million bytes/sec
client1
client2
2 50 GB 106.9 million bytes/sec
97.9 million bytes/sec
75.8 million bytes/sec
111.4 million bytes/sec
96.0 million bytes/sec
68.9 million bytes/sec
client1
client2
client3
1 50 GB 63.5 million bytes/sec 78.0 million bytes/sec client1
1 50 GB 69.4 million bytes/sec
68.1 million bytes/sec
73.4 million bytes/sec
64.5 million bytes/sec
client1
client2
1 50 GB 66.0 million bytes/sec
60.0 million bytes/sec
50.9 million bytes/sec
59.6 million bytes/sec
54.2 million bytes/sec
51.0 million bytes/sec
client1
client2
client3
3 33 GB 176.9 million bytes/sec 167.4 million bytes/sec client1
3 33 GB 155.6 million bytes/sec
142.1 million bytes/sec
150.6 million bytes/sec
154.5 million bytes/sec
client1
client2
3 33 GB 148.3 million bytes/sec
143.3 million bytes/sec
99.4 million bytes/sec
149.7 million bytes/sec
132.4 million bytes/sec
105.0 million bytes/sec
client1
client2
client3
1 33 GB 77.7 million bytes/sec 75.9 million bytes/sec client1
1 33 GB 72.6 million bytes/sec
67.9 million bytes/sec
68.4 million bytes/sec
61.6 million bytes/sec
client1
client2
1 33 GB 62.2 million bytes/sec
52.7 million bytes/sec
43.6 million bytes/sec
59.0 million bytes/sec
44.2 million bytes/sec
56.7 million bytes/sec
client1
client2
client3
4 25 GB 202.6 million bytes/sec 199.6 million bytes/sec client1
4 25 GB 169.6 million bytes/sec
158.6 million bytes/sec
188.7 million bytes/sec
187,0 million bytes/sec
client1
client2
4 25 GB 186.1 million bytes/sec
158.4 million bytes/sec
105.9 million bytes/sec
176.9 million bytes/sec
160.0 million bytes/sec
110.8 million bytes/sec
client1
client2
client3
1 25 GB 115.2 million bytes/sec 112.3 million bytes/sec client1
1 25 GB 61.2 million bytes/sec
69.9 million bytes/sec
59.5 million bytes/sec
54.9 million bytes/sec
client1
client2
1 25 GB 67.9 million bytes/sec
50.8 million bytes/sec
40.1 million bytes/sec
40.0 million bytes/sec
65.5 million bytes/sec
55.8 million bytes/sec
client1
client2
client3
5 20 GB 239.1 million bytes/sec 235.5 million bytes/sec client1
5 20 GB 237.0 million bytes/sec
200.3 million bytes/sec
206.5 million bytes/sec
204.5 million bytes/sec
client1
client2
5 20 GB 176.0 million bytes/sec
178.4 million bytes/sec
109.0 million bytes/sec
202.6 million bytes/sec
159.0 million bytes/sec
126.8 million bytes/sec
client1
client2
client3
1 20 GB 73.5 million bytes/sec 75.9 million bytes/sec client1
1 20 GB 67.7 million bytes/sec
57.0 million bytes/sec
72.0 million bytes/sec
60.8 million bytes/sec
client1
client2
1 20 GB 58.4 million bytes/sec
54.6 million bytes/sec
53.2 million bytes/sec
65.9 million bytes/sec
49.6 million bytes/sec
60.2 million bytes/sec
client1
client2
client3
10 10 GB 278.9 million bytes/sec 266.4 million bytes/sec client1
10 10 GB 235.5 million bytes/sec
240.7 million bytes/sec
252.0 million bytes/sec
200.7 million bytes/sec
client1
client2
10 10 GB 168.6 million bytes/sec
191.1 million bytes/sec
123.8 million bytes/sec
174.3 million bytes/sec
192.4 million bytes/sec
123.7 million bytes/sec
client1
client2
client3
1 10 GB 79.0 million bytes/sec 73.5 million bytes/sec client1
1 10 GB 72.0 million bytes/sec
71.1 million bytes/sec
77.2 million bytes/sec
67.5 million bytes/sec
client1
client2
1 10 GB 67.1 million bytes/sec
63.9 million bytes/sec
71.1 million bytes/sec
58.7 million bytes/sec
64.7 million bytes/sec
55.6 million bytes/sec
client1
client2
client3
20 5 GB 309.4 million bytes/sec 307.7 million bytes/sec client1
20 5 GB 236.5 million bytes/sec
192.1 million bytes/sec
222.3 million bytes/sec
193.5 million bytes/sec
client1
client2
20 5 GB 237.0 million bytes/sec
155.6 million bytes/sec
159.5 million bytes/sec
182.9 million bytes/sec
204.1 million bytes/sec
155.2 million bytes/sec
client1
client2
client3
1 5 GB 82.6 million bytes/sec 86.6 million bytes/sec client1
1 5 GB 80.1 million bytes/sec
79.0 million bytes/sec
82.6 million bytes/sec
81.3 million bytes/sec
client1
client2
1 5 GB 31.8 million bytes/sec
17.7 million bytes/sec
82.6 million bytes/sec
19.9 million bytes/sec
36.3 million bytes/sec
85.2 million bytes/sec
client1
client2
client3
33 3 GB 296.9 million bytes/sec 292.0 million bytes/sec client1
33 3 GB 191.5 million bytes/sec
183.3 million bytes/sec
189.5 million bytes/sec
204.0 million bytes/sec
client1
client2
33 3 GB 162.5 million bytes/sec
151.9 million bytes/sec
113.7 million bytes/sec
177.5 million bytes/sec
172.0 million bytes/sec
128.9 million bytes/sec
client1
client2
client3
1 3 GB 82.6 million bytes/sec 87.0 million bytes/sec client1
1 3 GB 82.6 million bytes/sec
82.6 million bytes/sec
82.6 million bytes/sec
82.6 million bytes/sec
client1
client2
1 3 GB 52.8 million bytes/sec
19.9 million bytes/sec
52.8 million bytes/sec
25.0 million bytes/sec
48.1 million bytes/sec
54.6 million bytes/sec
client1
client2
client3
50 2 GB 320.5 million bytes/sec 314.0 million bytes/sec client1
50 2 GB 205.7 million bytes/sec
204.9 million bytes/sec
226.5 million bytes/sec
200.0 million bytes/sec
client1
client2
50 2 GB 200.0 million bytes/sec
198.1 million bytes/sec
134.7 million bytes/sec
198.5 million bytes/sec
192.8 million bytes/sec
119.0 million bytes/sec
client1
client2
client3
1 2 GB 82.6 million bytes/sec 85.9 million bytes/sec client1
1 2 GB 74.1 million bytes/sec
71.6 million bytes/sec
76.7 million bytes/sec
67.1 million bytes/sec
client1
client2
1 2 GB 41.3 million bytes/sec
27.9 million bytes/sec
74.1 million bytes/sec
22.8 million bytes/sec
74.1 million bytes/sec
71.6 million bytes/sec
client1
client2
client3
100 1 GB 328.4 million bytes/sec 317.7 million bytes/sec client1
100 1 GB 233.9 million bytes/sec
208.5 million bytes/sec
231.9 million bytes/sec
226.0 million bytes/sec
client1
client2
100 1 GB 157.2 million bytes/sec
155.2 million bytes/sec
115.8 million bytes/sec
176.9 million bytes/sec
175.7 million bytes/sec
133.7 million bytes/sec
client1
client2
client3
1 1 GB 82.6 million bytes/sec 82.6 million bytes/sec client1
1 1 GB 22.4 million bytes/sec
56.5 million bytes/sec
63.2 million bytes/sec
67.1 million bytes/sec
client1
client2
1 1 GB 38.3 million bytes/sec
21.0 million bytes/sec
53.7 million bytes/sec
24.4 million bytes/sec
25.6 million bytes/sec
51.1 million bytes/sec
client1
client2
client3
200 500 MB 330.8 million bytes/sec 321.7 million bytes/sec client1
200 500 MB 246.7 million bytes/sec
208.1 million bytes/sec
223.1 million bytes/sec
241.1 million bytes/sec
client1
client2
200 500 MB 178.0 million bytes/sec
173.6 million bytes/sec
112.8 million bytes/sec
167.2 million bytes/sec
164.1 million bytes/sec
136.9 million bytes/sec
client1
client2
client3
300 333 MB 296.8 million bytes/sec 295.8 million bytes/sec client1
300 333 MB 196.6 million bytes/sec
191.4 million bytes/sec
177.1 million bytes/sec
207.9 million bytes/sec
client1
client2
300 333 MB 146.3 million bytes/sec
162.4 million bytes/sec
104.0 million bytes/sec
231.9 million bytes/sec
149.1 million bytes/sec
122.2 million bytes/sec
client1
client2
client3
400 250 MB 323.6 million bytes/sec 317.8 million bytes/sec client1
400 250 MB 217.5 million bytes/sec
219.4 million bytes/sec
222.2 million bytes/sec
222.2 million bytes/sec
client1
client2
400 250 MB 275.9 million bytes/sec
176.5 million bytes/sec
115.4 million bytes/sec
279.6 million bytes/sec
166.4 million bytes/sec
132.9 million bytes/sec
client1
client2
client3
500 200 MB 319.7 million bytes/sec 316.8 million bytes/sec client1
500 200 MB 238.3 million bytes/sec
217.1 million bytes/sec
219.4 million bytes/sec
250.3 million bytes/sec
client1
client2
500 200 MB 297.9 million bytes/sec
177.7 million bytes/sec
113.7 million bytes/sec
279.6 million bytes/sec
167.2 million bytes/sec
133.7 million bytes/sec
client1
client2
client3
1,000 100 MB 482.5 million bytes/sec 475.2 million bytes/sec client1
1,000 100 MB 354.3 million bytes/sec
443.1 million bytes/sec
321.0 million bytes/sec
406.4 million bytes/sec
client1
client2
1,000 100 MB 440.6 million bytes/sec
266.6 million bytes/sec
170.6 million bytes/sec
428.6 million bytes/sec
254.5 million bytes/sec
201.9 million bytes/sec
client1
client2
client3

4. Mirage Directory Speed Tests

Directory processing on Mirage may require use of all processor cores. This is due to the nature of a directory where it may contain 1-n files stored within the root directory, and 1-n sub-directories. When a directory is passed into Mirage's input queue and approved for processing, the primary thread is used to control directory processing thus preventing any other activity from occurring on the primary thread. While other cores may be in use processing earlier files, the handling of processing a directory will use all available cores as new cores become available to process the entire directory chain as quickly as possible.

Our directory tests occurred on the Mirage server side, and also occurred from client1 to/from Mirage. We performed the server side tests to determine baseline processing speed. The client side test demonstrate the overhead needed to transfer a complete directory to Mirage before processing can begin. The transfer of the processed directory back to the client can only occur once the entire directory has been fully processed. This cycle repeats to decrypt a directory containing encrypted files.

Note that a directory can contain a mixture of plaintext and encrypted files, where passing this directory makeup to Mirage will result in the inversion of each file. Plaintext files will be encrypted, where encrypted files will be decrypted.

Server-side test results

Files Total Size Sub-directories Encryption Decryption
33 310.3 billion bytes 17 1.12 billion bytes/sec 954.8 million bytes/sec
17 164.3 billion bytes 17 1.40 billion bytes/sec 1.54 billion bytes/sec
12 83.8 billion bytes 12 1.25 billion bytes/sec 1.23 billion bytes/sec
4 10.7 billion bytes 4 511.3 million bytes/sec 511.3 million bytes/sec

Client-server test results

Files Total Size Sub-directories Encryption Decryption
33 310.3 billion bytes 17 107.7 million bytes/sec 108.2 million bytes/sec
17 164.3 billion bytes 17 116.1 million bytes/sec 114.8 million bytes/sec
12 83.8 billion bytes 12 112.0 million bytes/sec 112.4 million bytes/sec
4 10.7 billion bytes 4 100.3 million bytes/sec 97.6 million bytes/sec

Click this link to view the Kryptera Success Story that was published by CENGN on October 11, 2018.