Review - Veggiehost And Primary VPS

This is a side to side comparison review for the two VPS. Both of the plans I joined were low-end, unmanaged and cheap.

First, lets see the plans:

Veggiehost - 2 dual core Opteron @ 96MHz, 256MB guaranteed RAM, 24GB HDD, 300GB bandwidth, $13.95 / month

PrimaryVPS - 2 dual core Xeon @ 1500MHz, 192MB guaranteed RAM, 12GB HDD, 175GB bandwidth, $13.95 / month

The plan from PrimaryVPS is a little bit downgraded. While as they are offering opening and quarterly payment special offer, you can find 256MB/15GB plan at comparable price.

Now, lets see the standard unixbench result. Each VPS took two benchmarking, one in peak hour and one in non-peak hour:

VEGGIEHOST

PHP Code:

Start Benchmark Run: Fri Apr 27 22:18:37 HKT 2007TEST                                        BASELINE     RESULT      INDEXDhrystone 2 using register variables        376783.7   951163.3       25.2Double-Precision Whetstone                      83.1     1920.5      231.1Execl Throughput                               188.3      173.9        9.2File Copy 1024 bufsize 2000 maxblocks         2672.0     8716.0       32.6File Copy 256 bufsize 500 maxblocks           1077.0     1895.0       17.6File Read 4096 bufsize 8000 maxblocks        15382.0    46337.0       30.1Pipe Throughput                             111814.6    26523.6        2.4Pipe-based Context Switching                 15448.6     7468.6        4.8Process Creation                               569.3      508.5        8.9Shell Scripts (8 concurrent)                    44.8       40.8        9.1System Call Overhead                        114433.5    46879.2        4.1                                                                 =========     FINAL SCORE                                                      13.8Start Benchmark Run: Mon Apr 30 08:09:55 HKT 2007Dhrystone 2 using register variables        376783.7   950332.3       25.2Double-Precision Whetstone                      83.1     1900.7      228.7Execl Throughput                               188.3      177.7        9.4File Copy 1024 bufsize 2000 maxblocks         2672.0     3576.0       13.4File Copy 256 bufsize 500 maxblocks           1077.0      974.0        9.0File Read 4096 bufsize 8000 maxblocks        15382.0    22736.0       14.8Pipe Throughput                             111814.6    24180.3        2.2Pipe-based Context Switching                 15448.6     7454.4        4.8Process Creation                               569.3      535.8        9.4Shell Scripts (8 concurrent)                    44.8       44.7       10.0System Call Overhead                        114433.5    47260.9        4.1                                                                 =========     FINAL SCORE                                                      11.3 

PRIMARYVPS

PHP Code:

Start Benchmark Run: Thu May 10 22:12:25 HKT 2007 Dhrystone 2 using register variables        376783.7  5775633.2      153.3Double-Precision Whetstone                      83.1      900.7      108.4Execl Throughput                               188.3     1665.1       88.4File Copy 1024 bufsize 2000 maxblocks         2672.0    64448.0      241.2File Copy 256 bufsize 500 maxblocks           1077.0    19882.0      184.6File Read 4096 bufsize 8000 maxblocks        15382.0   284552.0      185.0Pipe Throughput                             111814.6   527467.7       47.2Pipe-based Context Switching                 15448.6   101403.6       65.6Process Creation                               569.3     4162.2       73.1Shell Scripts (8 concurrent)                    44.8      341.8       76.3System Call Overhead                        114433.5   622391.7       54.4                                                                 =========     FINAL SCORE                                                     101.2PRIMARYVPS #2Start Benchmark Run: Mon May 14 09:38:14 HKT 2007 Dhrystone 2 using register variables        376783.7  3725455.0       98.9Double-Precision Whetstone                      83.1      897.0      107.9Execl Throughput                               188.3     1170.3       62.2File Copy 1024 bufsize 2000 maxblocks         2672.0    51839.0      194.0File Copy 256 bufsize 500 maxblocks           1077.0    19404.0      180.2File Read 4096 bufsize 8000 maxblocks        15382.0   411673.0      267.6Pipe Throughput                             111814.6   497630.1       44.5Pipe-based Context Switching                 15448.6    64217.7       41.6Process Creation                               569.3     5718.9      100.5Shell Scripts (8 concurrent)                    44.8      376.7       84.1System Call Overhead                        114433.5   724635.8       63.3                                                                 =========     FINAL SCORE                                                      95.7 

There are few findings from the result:

1. PRIMARYVPS is faster in overall result from CPU, IO to OS. While I could not believe that some results from VEGGIEHOST is even slower than a P2-233 BASELINE machine.

2. Though VEGGIEHOST is slower, its performance is VERY STABLE in both peak hour and non-peak hour. This is because VEGGIEHOST is now using a "hard limit" to ensure its "equal share" CPU policy since late March. This was confirmed by their support once I encountered performance drop for half in April.

Next test is using Apache Benchmark (ab) to test the theoretical webpages processing limit and also see the network throughput (from Hong Kong) in peak hour:

VEGGIEHOST

PHP Code:

1. Simple HTML static page   - About 300pg/s2. Simple LAMP dynamic page  - About 50pg/s3. Complex LAMP dynamic page - About 5pg/s4. Longest total time for #1 - About 800ms5. Longest total time for #3 - About 2000ms6. Throughput from remote    - About 64KBPS (+/- 2KBPS) 

PRIMARYVPS

PHP Code:

1. Simple HTML static page   - About 800pg/s2. Simple LAMP dynamic page  - About 130pg/s3. Complex LAMP dynamic page - About 13pg/s4. Longest total time for #1 - About 10ms5. Longest total time for #3 - About 600ms6. Throughput from remote    - About 56KBPS (+/- 2KBPS) 

The most interesting thing from the Apache bench is that the worst case to process the web pages from VEGGIEHOST is highly fluctuating. This suspected to be due to the limitation of IO partitioning in VPS technology -- An issue affecting performance much in VPS while not yet fully solved.

When check closer for IO performance from unixbench, we have the following results:

PURE DISK ACCESS

PHP Code:

Sun V480/SE3320 RAID5# ./fsbuffer 10 ; ./fsdisk 10 ; ./fstime 1013850 Kbytes/sec write 256 bufsize 2000 max blocks48567 Kbytes/sec read 256 bufsize 2000 max blocks10327 Kbytes/sec copy 256 bufsize 2000 max blocks36000 Kbytes/sec write 4096 bufsize 2000 max blocks419994 Kbytes/sec read 4096 bufsize 2000 max blocks29000 Kbytes/sec write 1024 bufsize 2000 max blocks138274 Kbytes/sec read 1024 bufsize 2000 max blocks25303 Kbytes/sec copy 1024 bufsize 2000 max blocksSun V480/INTERNAL RAID5# ./fsbuffer 10 ; ./fsdisk 10 ; ./fstime 1054000 Kbytes/sec write 256 bufsize 2000 max blocks52872 Kbytes/sec read 256 bufsize 2000 max blocks24670 Kbytes/sec copy 256 bufsize 2000 max blocks411200 Kbytes/sec write 4096 bufsize 2000 max blocks427934 Kbytes/sec read 4096 bufsize 2000 max blocks175800 Kbytes/sec write 1024 bufsize 2000 max blocks154007 Kbytes/sec read 1024 bufsize 2000 max blocks77536 Kbytes/sec copy 1024 bufsize 2000 max blocksVeggie Host# ./fsbuffer 10 ; ./fsdisk 10 ; ./fstime 108850 Kbytes/sec write 256 bufsize 2000 max blocks15018 Kbytes/sec read 256 bufsize 2000 max blocks5387 Kbytes/sec copy 256 bufsize 2000 max blocks54400 Kbytes/sec write 4096 bufsize 2000 max blocks88029 Kbytes/sec read 4096 bufsize 2000 max blocks30400 Kbytes/sec write 1024 bufsize 2000 max blocks49496 Kbytes/sec read 1024 bufsize 2000 max blocks18038 Kbytes/sec copy 1024 bufsize 2000 max blocksPrimary Vps# ./fsbuffer 10 ; ./fsdisk 10 ; ./fstime 1078850 Kbytes/sec write 256 bufsize 2000 max blocks157190 Kbytes/sec read 256 bufsize 2000 max blocks51128 Kbytes/sec copy 256 bufsize 2000 max blocks444000 Kbytes/sec write 4096 bufsize 2000 max blocks1115200 Kbytes/sec read 4096 bufsize 2000 max blocks215200 Kbytes/sec write 1024 bufsize 2000 max blocks505957 Kbytes/sec read 1024 bufsize 2000 max blocks167760 Kbytes/sec copy 1024 bufsize 2000 max blocks 

Though RAID5 is slow, I did not realize before that a VPS can outperform a dedicated server!

Till now it seems all the facts from Primary VPS is excellent, while if you want to join, please consider if the following issues would be a limitation to you:

1. VEGGIEHOST is using a subscription based payment, which means you can terminate the plan once no longer needed. On the other hand, PrimaryVPS requires 30 days prior to termination.

2. VEGGIEHOST is far relax in VPS usage while PrimaryVPS restrict tighter. You can burst the CPU at VEGGIEHOST for 7x24 since it applied a hard limit (in this case hard limit is a good thing) while PrimaryVPS gives you a much higher burst CPU limit but the terms is that you cannot burst over 15% (?) CPU time for more than 5 minutes in any case. Also, VEGGIEHOST does not limit whatever purposes for your VPS while PrimaryVPS do not allow something like P2P in your VPS. This can ensure the system (particularly the IO) did not being pulled by some applications, which is good for those who do not run the P2P applications, while may not be suitable if you need to run it.

3. I have used VEGGIEHOST for 2 months and quite certain that the above results are reliable, while only used PrimaryVPS for 1 week, and since it is a new provider I could not tell if the box will be degraded when more participants to join later. I will post updated review few weeks later to see if this issue occurred.