All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.
[NL] HostDZire vs Leaseweb VPS benchmark - differences
I got my hands on one of @HostDZire's own (own HW) VPS in NL, AMS and of course benchmarked it.
Here are the results (based on > 30 runs). As usual first sysinfo processor and memory.
Version 2.5.0a, (c) 2018+ jsg (->lowendtalk.com)
Machine: amd64, Arch.: amd64, Model: AMD EPYC 7K62 48-Core Processor
OS, version: FreeBSD 14.2, Mem.: 5.989 GB
CPU - Cores: 4, Family/Model/Stepping: 23/49/0
Cache: 32K/32K L1d/L1i, 512K L2, 192M L3
Std. Flags: fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat
pse36 cflsh mmx fxsr sse sse2 htt sse3 pclmulqdq ssse3 fma cx16
sse4_1 sse4_2 popcnt aes xsave osxsave avx f16c rdrnd hypervisor
Ext. Flags: syscall nx mmxext fxsr_opt pdpe1gb rdtscp lm lahf_lm cmp_legacy
extapic cr8_legacy lzcnt sse4a misalignsse 3dnowprefetch osvw
AES? Yes
Nested Virt.? No
HW RNG? Yes
ProcMem SC [MB/s]: avg 229.4 - min 92.8 (40.5 %), max 366.0 (159.6 %)
ProcMem MA [MB/s]: avg 990.3 - min 879.4 (88.8 %), max 1110.0 (112.1 %)
ProcMem MB [MB/s]: avg 1065.2 - min 968.5 (90.9 %), max 1190.0 (111.7 %)
ProcMem AES [MB/s]: avg 1151.5 - min 1144.2 (99.4 %), max 1157.4 (100.5 %)
ProcMem RSA [kp/s]: avg 101.1 - min 92.8 (91.8 %), max 107.8 (106.6 %)
Nice and normal results for an Epyc. Crypto results also nice. All desirable flags are there. The only point of potential lack of happiness: no embedded virtualization, which I normally do not care about at all but with 4 vCores it might not be nonsensical to embed one or more smaller VPSs within a relatively beefy VPS.
Now, let's look at the disk.
--- Disk 4 KB - Buffered ---
Write seq. [MB/s]: avg 11.17 - min 8.50 (76.1%), max 12.69 (113.7%)
Write rnd. [MB/s]: avg 10.91 - min 8.56 (78.5%), max 13.26 (121.5%)
Read seq. [MB/s]: avg 26.10 - min 18.67 (71.5%), max 34.72 (133.0%)
Read rnd. [MB/s]: avg 20.29 - min 14.51 (71.5%), max 30.75 (151.5%)
--- Disk 4 KB - Sync/Direct ---
Write seq. [MB/s]: avg 10.90 - min 9.25 (84.8%), max 13.38 (122.7%)
Write rnd. [MB/s]: avg 10.85 - min 9.50 (87.5%), max 13.41 (123.5%)
Read seq. [MB/s]: avg 26.36 - min 18.86 (71.6%), max 34.47 (130.8%)
Read rnd. [MB/s]: avg 23.62 - min 17.10 (72.4%), max 29.68 (125.6%)
--- Disk 64 KB - Buffered ---
Write seq. [MB/s]: avg 105.68 - min 89.90 (85.1%), max 123.18 (116.6%)
Write rnd. [MB/s]: avg 118.37 - min 99.80 (84.3%), max 139.77 (118.1%)
Read seq. [MB/s]: avg 2707.48 - min 2193.75 (81.0%), max 3538.00 (130.7%)
Read rnd. [MB/s]: avg 326.68 - min 216.00 (66.1%), max 424.00 (129.8%)
--- Disk 64 KB - Sync/Direct ---
Write seq. [MB/s]: avg 11.99 - min 9.72 (81.1%), max 13.14 (109.6%)
Write rnd. [MB/s]: avg 6.41 - min 5.41 (84.4%), max 6.88 (107.3%)
Read seq. [MB/s]: avg 2398.59 - min 1752.44 (73.1%), max 2857.05 (119.1%)
Read rnd. [MB/s]: avg 304.45 - min 230.67 (75.8%), max 380.80 (125.1%)
--- Disk 1 MB - Buffered ---
Write seq. [MB/s]: avg 172.89 - min 149.98 (86.7%), max 189.32 (109.5%)
Write rnd. [MB/s]: avg 354.21 - min 300.21 (84.8%), max 397.65 (112.3%)
Read seq. [MB/s]: avg 4418.78 - min 3775.67 (85.4%), max 5013.70 (113.5%)
Read rnd. [MB/s]: avg 1622.05 - min 1390.84 (85.7%), max 1891.14 (116.6%)
--- Disk 1 MB - Sync/Direct ---
Write seq. [MB/s]: avg 41.49 - min 35.17 (84.8%), max 43.94 (105.9%)
Write rnd. [MB/s]: avg 26.81 - min 22.54 (84.1%), max 28.09 (104.8%)
Read seq. [MB/s]: avg 3701.02 - min 3179.04 (85.9%), max 4109.73 (111.0%)
Read rnd. [MB/s]: avg 1696.15 - min 1420.89 (83.8%), max 1987.98 (117.2%)
--- Disk IOps (Sync/Direct) ---
Write seq. [MB/s]: avg 36.55 - min 34.62 (94.7%), max 37.90 (103.7%)
IOps : avg 9357.70 - min 8861.50 (94.7%), max 9701.23 (103.7%)
Wow, that's a quite nice NVMe with close to 10000 IOps!
(Side note: while I do not know how many nodes with RAID'ed disk @HostDZire already have there, I know that this VPS has such a disk and there seem to be plans to upgrade all nodes)
Finally connectivity, which I expect to be quite decent as the Leaseweb network is used.
--- Europe ---
NO OSL mirror.terrahost.no [F: 0]
DL [Mb/s]: avg 479.3 - min 90.6 (18.9%), max 609.3 (127.1%)
Ping [ms]: avg 20.0 - min 17.3 (86.5%), max 21.1 (105.5%)
Web ping [ms]: avg 61.3 - min 17.5 (28.5%), max 128.0 (208.7%)
UK LON lon.speedtest.clouvider.net [F: 0]
DL [Mb/s]: avg 1058.8 - min 166.5 (15.7%), max 1471.7 (139.0%)
Ping [ms]: avg 8.8 - min 8.6 (97.2%), max 10.3 (116.5%)
Web ping [ms]: avg 8.9 - min 8.6 (96.6%), max 10.3 (115.7%)
NL AMS nl.mirrors.clouvider.net [F: 0]
DL [Mb/s]: avg 2455.6 - min 1500.9 (61.1%), max 3202.3 (130.4%)
Ping [ms]: avg 1.0 - min 0.9 (90.9%), max 1.3 (131.4%)
Web ping [ms]: avg 1.3 - min 0.9 (69.2%), max 2.1 (161.5%)
DE FRA mirror.plusline.net [F: 0]
DL [Mb/s]: avg 1190.4 - min 369.8 (31.1%), max 1395.9 (117.3%)
Ping [ms]: avg 6.7 - min 6.6 (99.1%), max 7.3 (109.6%)
Web ping [ms]: avg 7.8 - min 6.8 (86.9%), max 22.7 (289.9%)
FR PAR ftp1.fr.freebsd.org [F: 0]
DL [Mb/s]: avg 554.5 - min 98.0 (17.7%), max 629.1 (113.5%)
Ping [ms]: avg 11.7 - min 11.4 (97.8%), max 14.6 (125.3%)
Web ping [ms]: avg 33.2 - min 11.4 (34.4%), max 123.2 (371.5%)
CH ZUR mirror.init7.net [F: 0]
DL [Mb/s]: avg 736.9 - min 147.3 (20.0%), max 852.4 (115.7%)
Ping [ms]: avg 12.5 - min 12.4 (99.1%), max 12.7 (101.5%)
Web ping [ms]: avg 14.8 - min 12.6 (85.4%), max 17.7 (120.0%)
ES MAD mirror.raiolanetworks.com [F: 0]
DL [Mb/s]: avg 425.9 - min 365.0 (85.7%), max 447.7 (105.1%)
Ping [ms]: avg 25.9 - min 24.7 (95.5%), max 30.7 (118.7%)
Web ping [ms]: avg 27.9 - min 24.9 (89.3%), max 35.1 (125.8%)
RO BUC mirrors.hosterion.ro [F: 0]
DL [Mb/s]: avg 298.0 - min 287.8 (96.6%), max 306.8 (103.0%)
Ping [ms]: avg 34.9 - min 0.0 (0.0%), max 36.5 (104.5%)
Web ping [ms]: avg 35.2 - min 0.0 (0.0%), max 38.0 (108.1%)
RU MOS speedtest.hostkey.ru [F: 0]
DL [Mb/s]: avg 253.3 - min 39.7 (15.7%), max 290.1 (114.5%)
Ping [ms]: avg 40.6 - min 40.4 (99.6%), max 40.9 (100.8%)
Web ping [ms]: avg 61.5 - min 42.9 (69.8%), max 303.1 (493.0%)
--- Asia / Oceania ---
RU SIB mirror.truenetwork.ru [F: 2]
DL [Mb/s]: avg 117.0 - min 0.0 (0.0%), max 132.7 (113.4%)
Ping [ms]: avg 81.3 - min 0.0 (0.0%), max 83.6 (102.8%)
Web ping [ms]: avg 94.4 - min 0.0 (0.0%), max 168.0 (177.9%)
IN MUM mirrors.piconets.webwerks.in [F: 0]
DL [Mb/s]: avg 44.9 - min 29.4 (65.6%), max 48.0 (106.8%)
Ping [ms]: avg 253.9 - min 0.0 (0.0%), max 268.8 (105.9%)
Web ping [ms]: avg 257.4 - min 0.0 (0.0%), max 283.3 (110.0%)
SG SGP mirror.sg.gs [F: 0]
DL [Mb/s]: avg 69.3 - min 61.2 (88.3%), max 71.5 (103.1%)
Ping [ms]: avg 161.1 - min 159.1 (98.8%), max 163.5 (101.5%)
Web ping [ms]: avg 164.5 - min 160.1 (97.3%), max 168.0 (102.1%)
CN HKG mirrors.xtom.hk [F: 0]
DL [Mb/s]: avg 56.9 - min 43.6 (76.6%), max 62.0 (108.9%)
Ping [ms]: avg 184.9 - min 184.7 (99.9%), max 185.0 (100.1%)
Web ping [ms]: avg 193.3 - min 184.8 (95.6%), max 297.9 (154.1%)
CN BEJ mirrors.bfsu.edu.cn [F: 0]
DL [Mb/s]: avg 54.5 - min 35.1 (64.4%), max 66.5 (122.0%)
Ping [ms]: avg 232.0 - min 0.0 (0.0%), max 250.5 (108.0%)
Web ping [ms]: avg 232.2 - min 0.0 (0.0%), max 250.5 (107.9%)
JP OSA mirrors.xtom.jp [F: 0]
DL [Mb/s]: avg 43.9 - min 39.0 (88.7%), max 45.2 (103.1%)
Ping [ms]: avg 250.8 - min 250.6 (99.9%), max 251.3 (100.2%)
Web ping [ms]: avg 251.2 - min 250.6 (99.8%), max 254.1 (101.2%)
AU SYD mirror.internet.asn.au [F: 1]
DL [Mb/s]: avg 44.3 - min 0.0 (0.0%), max 46.8 (105.6%)
Ping [ms]: avg 251.9 - min 250.9 (99.6%), max 266.3 (105.7%)
Web ping [ms]: avg 252.0 - min 251.0 (99.6%), max 266.3 (105.7%)
--- Africa ---
KE NAI mirror.liquidtelecom.com [F: 11]
DL [Mb/s]: avg 27.0 - min 0.0 (0.0%), max 61.4 (227.3%)
Ping [ms]: avg 138.9 - min 0.0 (0.0%), max 210.6 (151.6%)
Web ping [ms]: avg 158.9 - min 0.0 (0.0%), max 1015.3 (638.8%)
ZA WEC archlinux.za.mirror.allworldit.com [F: 0]
DL [Mb/s]: avg 63.1 - min 25.7 (40.7%), max 71.9 (113.9%)
Ping [ms]: avg 156.7 - min 0.0 (0.0%), max 185.2 (118.2%)
Web ping [ms]: avg 179.6 - min 0.0 (0.0%), max 215.8 (120.1%)
--- Americas ---
CA MTL speedtest.mtl2.ca.leaseweb.net [F: 0]
DL [Mb/s]: avg 125.3 - min 25.3 (20.2%), max 139.4 (111.3%)
Ping [ms]: avg 82.2 - min 81.2 (98.8%), max 93.9 (114.3%)
Web ping [ms]: avg 82.5 - min 81.4 (98.6%), max 93.9 (113.8%)
US NYC nyc.mirrors.clouvider.net [F: 0]
DL [Mb/s]: avg 132.6 - min 42.8 (32.3%), max 141.5 (106.7%)
Ping [ms]: avg 77.5 - min 77.0 (99.4%), max 81.5 (105.2%)
Web ping [ms]: avg 77.8 - min 77.2 (99.2%), max 81.7 (105.0%)
US ASH ash.speedtest.clouvider.net [F: 0]
DL [Mb/s]: avg 115.2 - min 105.9 (92.0%), max 128.4 (111.5%)
Ping [ms]: avg 97.9 - min 97.7 (99.8%), max 98.7 (100.8%)
Web ping [ms]: avg 98.0 - min 97.7 (99.7%), max 98.7 (100.7%)
US PIB mirror.pit.teraswitch.com [F: 0]
DL [Mb/s]: avg 113.5 - min 32.4 (28.6%), max 121.5 (107.1%)
Ping [ms]: avg 92.3 - min 92.1 (99.8%), max 92.8 (100.5%)
Web ping [ms]: avg 93.8 - min 92.2 (98.3%), max 97.3 (103.7%)
US MIA speedtest.mia11.us.leaseweb.net [F: 0]
DL [Mb/s]: avg 95.3 - min 33.1 (34.7%), max 108.9 (114.3%)
Ping [ms]: avg 105.1 - min 104.8 (99.7%), max 106.3 (101.1%)
Web ping [ms]: avg 105.6 - min 104.9 (99.3%), max 107.2 (101.5%)
US CHI ord.mirror.rackspace.com [F: 0]
DL [Mb/s]: avg 121.4 - min 113.8 (93.7%), max 124.5 (102.5%)
Ping [ms]: avg 93.5 - min 93.3 (99.8%), max 93.7 (100.2%)
Web ping [ms]: avg 128.4 - min 93.6 (72.9%), max 176.8 (137.7%)
US ATL atl.speedtest.clouvider.net [F: 0]
DL [Mb/s]: avg 106.2 - min 36.0 (33.9%), max 119.3 (112.4%)
Ping [ms]: avg 102.7 - min 101.0 (98.3%), max 104.3 (101.6%)
Web ping [ms]: avg 103.5 - min 101.0 (97.6%), max 117.7 (113.8%)
US PHO phx.speedtest.clouvider.net [F: 0]
DL [Mb/s]: avg 77.3 - min 20.9 (27.1%), max 85.5 (110.5%)
Ping [ms]: avg 130.2 - min 128.8 (98.9%), max 132.4 (101.7%)
Web ping [ms]: avg 131.5 - min 129.1 (98.2%), max 136.4 (103.7%)
US PTL mirrors.cat.pdx.edu [F: 3]
DL [Mb/s]: avg 66.9 - min 0.0 (0.0%), max 77.3 (115.5%)
Ping [ms]: avg 151.9 - min 148.7 (97.9%), max 153.2 (100.9%)
Web ping [ms]: avg 154.0 - min 149.0 (96.8%), max 160.8 (104.4%)
US LAX speedtest.lax12.us.leaseweb.net [F: 0]
DL [Mb/s]: avg 82.0 - min 26.3 (32.1%), max 87.0 (106.2%)
Ping [ms]: avg 131.8 - min 130.8 (99.3%), max 134.0 (101.7%)
Web ping [ms]: avg 132.8 - min 131.2 (98.8%), max 138.8 (104.5%)
US SJO mirrors.xtom.us [F: 0]
DL [Mb/s]: avg 78.7 - min 63.5 (80.6%), max 82.4 (104.7%)
Ping [ms]: avg 140.3 - min 140.2 (99.9%), max 140.8 (100.3%)
Web ping [ms]: avg 140.5 - min 140.2 (99.8%), max 142.2 (101.2%)
US SEA speedtest.sea11.us.leaseweb.net [F: 0]
DL [Mb/s]: avg 71.9 - min 26.9 (37.4%), max 84.0 (116.8%)
Ping [ms]: avg 139.8 - min 137.2 (98.2%), max 140.8 (100.7%)
Web ping [ms]: avg 139.9 - min 137.2 (98.1%), max 141.6 (101.2%)
BR SPA mirror.ufscar.br [F: 2]
DL [Mb/s]: avg 45.1 - min 0.0 (0.0%), max 54.2 (120.3%)
Ping [ms]: avg 210.1 - min 208.6 (99.3%), max 211.6 (100.7%)
Web ping [ms]: avg 220.8 - min 209.8 (95.0%), max 252.6 (114.4%)
BR CUR archlinux.c3sl.ufpr.br [F: 0]
DL [Mb/s]: avg 46.7 - min 22.3 (47.7%), max 52.6 (112.6%)
Ping [ms]: avg 211.8 - min 210.0 (99.1%), max 251.9 (118.9%)
Web ping [ms]: avg 221.6 - min 210.2 (94.9%), max 264.3 (119.3%)
CL SAN elmirror.cl [F: 3]
DL [Mb/s]: avg 45.2 - min 0.0 (0.0%), max 54.4 (120.3%)
Ping [ms]: avg 217.2 - min 216.2 (99.5%), max 237.8 (109.5%)
Web ping [ms]: avg 225.5 - min 216.2 (95.9%), max 267.1 (118.4%)
Europe - Very nice indeed. And a few results above 1 Gb/s which confirms that the node has a 10 Gb/s pipe (or 5 Gb/s but I guess professional providers like Leaseweb don't play around but go directly from 1 Gb/s to 10 Gb/s.
Asia / Oceania - Hmmm, IN, Mumbai quite weak, SGP and HongKong not really strong as well but some quite fast China targets. Japan and Straya also mediocre.
Let me put it like this: if good Asia / Oceania connectivity is what you need I'd look elsewhere.
Africa - half-cooked. One target quite decent, the other quite poor.
Americas - When I see east-coast targets at about 125 Mb/s and Chicago at about 120 Mb/s (from Europe) I know I'm looking at a winner wrt cross-Atlantic connectivity. California at about 80 Mb/s is quite decent as well. What I also find really noteworthy are decent results on both south-american coasts.
Summary/verdict: a really nice VPS with excellent connectivity within Europe and to North-America. Asia / Oceania is the only bitter pill wrt connectivity.
Clearly recommended and even a cheapie for what you get (as long as you can live with mediocre Asia / Oceania connectivity and a bit of risk wrt your data, until the upgrade is complete for all nodes).
But there's also the Leaseweb "original" version!
And there were a few discussions recently about which version is more attractive, which triggered this examination and thread.
To avoid misunderstandings: based on what I know both products (HD and LW VPS) use the same colo and network. HostDZire's product has/runs on its own hardware however - *that's it, that's the difference (modulo temporarily some non-RAID disks).
Well, the first thing that triggered my interest is the price difference and AFAIK it's hefty ($47/yr for LW, $26/yr for HD IIRC).
So obviously the question arises what morer or betterer do you get for almost double the $$ (from LW)?
Let's have a good close look. And yes, of course I also benchmarked the Leaseweb version!
Here's the results, based on > 30 runs.
Version 2.5.0a, (c) 2018+ jsg (->lowendtalk.com)
Machine: amd64, Arch.: amd64, Model: AMD EPYC Processor
OS, version: FreeBSD 14.4, Mem.: 5.989 GB
CPU - Cores: 4, Family/Model/Stepping: 23/1/2
Cache: 32K/64K L1d/L1i, 512K L2, 8M L3
Std. Flags: fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat
pse36 cflsh mmx fxsr sse sse2 htt sse3 pclmulqdq ssse3 fma cx16
sse4_1 sse4_2 popcnt aes xsave osxsave avx f16c rdrnd hypervisor
Ext. Flags: syscall nx mmxext fxsr_opt pdpe1gb rdtscp lm lahf_lm cmp_legacy
cr8_legacy lzcnt sse4a misalignsse 3dnowprefetch osvw topoext
AES? Yes
Nested Virt.? No
HW RNG? Yes
ProcMem SC [MB/s]: avg 236.6 - min 93.1 (39.4 %), max 383.1 (161.9 %)
ProcMem MA [MB/s]: avg 806.8 - min 574.6 (71.2 %), max 1070.0 (132.6 %)
ProcMem MB [MB/s]: avg 953.4 - min 673.5 (70.6 %), max 1120.0 (117.5 %)
ProcMem AES [MB/s]: avg 1133.7 - min 1059.3 (93.4 %), max 1179.2 (104.0 %)
ProcMem RSA [kp/s]: avg 101.9 - min 93.1 (91.4 %), max 109.0 (107.0 %)
Very little difference at first glance.
But there's an interesting detail: Single core performance on Leaseweb is (slightly) better while multi-core performance is better on the HostDZire version. Interesting. How come?
Hint: Look at the L1 and L3 caches of both! LW has a larger L1i cache (64 KB vs 32 KB) which translates to faster single-core performance, while HD has a giant L3 cache, which leads to higher multi-threading performance.
I personally prefer HostDZire's approach. Simple reason: A VPS node almost always has many VPSs and with quite different workloads and lots and lots of task switches. L2 can't help a whole lot if there are more than 8 VPS on a node (512KB div 64KB = 8 (graciously ignoring the 32KB L1d)) and that view still is almost idiotically in favour of LW because it assumes that each VPS does exactly the same task, i.e. internal task switches don't invalidate L1.
A massive L3 cache though can and does buffer a significant parts of (very time expensive) DRAM accesses which significantly benefits VPS nodes (as the results show).
processor & memory summary: HostDZire's processor choice is clearly (and demonstrably) better.
Now, let's look at the disk.
--- Disk 4 KB - Buffered ---
Write seq. [MB/s]: avg 8.25 - min 7.68 (93.0%), max 9.17 (111.1%)
Write rnd. [MB/s]: avg 8.12 - min 7.70 (94.9%), max 8.68 (106.9%)
Read seq. [MB/s]: avg 23.84 - min 22.07 (92.6%), max 25.06 (105.1%)
Read rnd. [MB/s]: avg 21.20 - min 19.76 (93.2%), max 23.20 (109.4%)
--- Disk 4 KB - Sync/Direct ---
Write seq. [MB/s]: avg 8.23 - min 7.75 (94.1%), max 8.87 (107.7%)
Write rnd. [MB/s]: avg 8.09 - min 7.48 (92.5%), max 8.54 (105.6%)
Read seq. [MB/s]: avg 23.74 - min 22.27 (93.8%), max 25.05 (105.5%)
Read rnd. [MB/s]: avg 21.13 - min 19.00 (89.9%), max 22.39 (105.9%)
--- Disk 64 KB - Buffered ---
Write seq. [MB/s]: avg 76.18 - min 59.20 (77.7%), max 83.04 (109.0%)
Write rnd. [MB/s]: avg 83.59 - min 79.16 (94.7%), max 88.20 (105.5%)
Read seq. [MB/s]: avg 2846.76 - min 2627.32 (92.3%), max 3006.36 (105.6%)
Read rnd. [MB/s]: avg 294.03 - min 269.53 (91.7%), max 312.96 (106.4%)
--- Disk 64 KB - Sync/Direct ---
Write seq. [MB/s]: avg 7.75 - min 6.46 (83.4%), max 9.24 (119.3%)
Write rnd. [MB/s]: avg 4.16 - min 3.59 (86.2%), max 4.87 (117.0%)
Read seq. [MB/s]: avg 2829.59 - min 2594.43 (91.7%), max 3082.77 (108.9%)
Read rnd. [MB/s]: avg 292.52 - min 271.14 (92.7%), max 310.11 (106.0%)
--- Disk 1 MB - Buffered ---
Write seq. [MB/s]: avg 111.54 - min 76.28 (68.4%), max 119.04 (106.7%)
Write rnd. [MB/s]: avg 218.73 - min 201.14 (92.0%), max 237.75 (108.7%)
Read seq. [MB/s]: avg 4374.31 - min 4199.96 (96.0%), max 4641.00 (106.1%)
Read rnd. [MB/s]: avg 1342.01 - min 1087.97 (81.1%), max 1563.24 (116.5%)
--- Disk 1 MB - Sync/Direct ---
Write seq. [MB/s]: avg 27.23 - min 24.82 (91.2%), max 29.53 (108.5%)
Write rnd. [MB/s]: avg 17.60 - min 16.23 (92.2%), max 18.70 (106.2%)
Read seq. [MB/s]: avg 4383.68 - min 4141.23 (94.5%), max 4547.53 (103.7%)
Read rnd. [MB/s]: avg 1348.34 - min 1217.51 (90.3%), max 1622.98 (120.4%)
--- Disk IOps (Sync/Direct) ---
Write seq. [MB/s]: avg 27.52 - min 26.55 (96.5%), max 28.66 (104.2%)
IOps : avg 7044.35 - min 6797.06 (96.5%), max 7337.33 (104.2%)
HostDZire's disk clearly is significantly faster - but: Leaseweb's disk is roughly in the same ballpark and by no means crappy. Plus, albeit only temporary, Leaseweb's disk is RAID'ed while HostDZire hasn't completed their upgrade process yet (AFAIK).
TL;DR / verdict
I'll put it like this: as I'm not running critical business stuff on that VPS (in fact on most of my VPS) I'm not really disturbed by not all of HD's nodes having RAID yet (mine is RAID'ed though, lucky me *g).
HostDZire's version IMO has a better processor & memory combo, a faster disk (again: modulo not yet RAID on all nodes), exactly the same connectivity as LW, and is in fact in the same colo as well.
And is way cheaper!
Don't get me wrong, I'm a big Leaseweb fan and certainly not inclined to bash them in any way! But at the end of the day I get more performance for a much lower price with the HostDZire VPS version.
So, I sincerely thank HostDZire for providing a LW VPS to me for benchmarking, but I'll give it back and keep my NL, AMS HostDZire VPS. It's better and cheaper, simple as that.
Comments
Nice analysis. When testing write, how big was the blob?
2 GB total in chunks. For the IOps numbers these are based on 4KB slices and 4 threads.
Uh, I see. Usually I try to use larger blob to exhaust the burst quota
@jsg Thannk you for the benchmark report. as always very helpful for us so we can find room to improve.
You're welcome, but I merely reported the facts I measured and some of my conclusions.
It might be useful though to give us an ETA re. when all nodes are RAID updated.