New on LowEndTalk? Please Register and read our Community Rules.
All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.
All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.
Comments
major update
"Why didn't you review all @HostDZire / Leaseweb locations?!" you ask? You are right, but hey, it takes time to do 30+ benchmark runs! That's why I publish this review update "only now".
What did I benchmark and do I review in this update? The UK, LAX, and SFO locations.
Let's start with LAX.
Well, the usual Leaseweb performance. Not super-high but very decent indeed. Also note the relatively low spread and the good multi-core to single-core ratio.
On to the disk
Well, again the usual quite good Leaseweb performance. Not high-end but really decent NVMe territory.
Being at that, it seems that Leaseweb chose a smart position in the provider race. First because they had a clear target group in mind, business customers who typically don't care that much about "racing" high-end performance, neither with the processor, nor with the disk; what they do care about though is (a) reliable and reliably decent performance, and (b) a very good and global network - and that's what they get from Leaseweb / HostDZire. If they need more processor performance they simply upgrade to a VPS with more cores and memory. And for really serious tasks they simply get a dedi.
But with their prices and HostDZire selling those VPS to us mere mortals we can get what actually is a really decent professional business VPS
So, let's look at the connectivity of that VPS
First a note: The usual few targets were not reachable but OTOH all secondary targets in Europe, CN, and Iran were reached and achieved decent results.
As usual I'll split it into continents/regions.
Europe - I (a European) am not particularly happy with those results, but OK, LAX is at the far end of the country and notoriously slow from Europe and, as we see now, vice versa as well.
Asia/Oceania - Again, not exactly great, but also again, far, far away. I guess all in all the results are about as good as one can get from the west-coast (modulo a few specialist providers/carriers). In case any american readers know better providers from Murrica to Asia / Oceanian I'd appreciate some info
Africa - meh
America - Funny. A LAX location with higher results to SFO than to another LAX location g. Solidly above 150 Mb/s to NYC, i.e. across the whole nation, is respectable I guess, as is 200+Mb/s to Chicago although I would have expected better. And over 1 Gb/s to SFO is really good, I guess (worded carefully because, again, my experience with connectivity there is quite limited).
TL;DR/verdict: A very decent typical Leaseweb VPS with, I guess, decent connectivity for a VPS in California.
Now, Let's have a look at the SFO VPS.
I'll skip sysinfo, processor, and memory as it's basically the same as the LAX VPS. So let's rather look at the strikingly similar disk results (not really a surprise).
Well, just read my comments re the disk above in the LAX review ...
On to the connectivity, maybe there we'll find differences that are not insignificant.
And again, split into continents/regions.
Europe - Pretty much the same as the LAX results. A few locations a bit faster, a few others a bit slower, but no really significant differences.
Asia / Oceania - Maybe a tad bit slower but basically no really significant difference again.
Africa - same, same again.
America - This is interesting (and also a bit weird). OK, NYC and CHI basically the same with SFO only a modest 10 - 15 Mb/s faster than the LAX VPS. But "locally" it gets interesting. SFO VPS to LAX is only less than 100 Mb/s slower than LAX VPS to LAX! But the SFO VPS achieves an almost 4 x higher result to San Jose!
TL;DR/verdict: the LAX VPS and the SFO VPS basically are twin sisters. Basically no difference at all wrt processor, memory, and disk and only neglegible to minimal differences in connectivity - with one very significant exception: The SFO VPS is very significantly faster within California.
All in all I'll put it like this: I'm glad that I purchased the WDC promo VPS, because it's significantly better then a the west-coast VPSs re connectivity to Europe, and slow Asia routes I have plenty in Europe *g
Finally to the UK VPS ...
I'll again skip sysinfo, processor, and memory because it's just yet another Leaseweb "cloned" node.
So, let's have a look at the disk.
Yet again basically just another Leaseweb disk "clone", albeit this one is a bee stick slower.
So let's go at the beef right away, to the connectivity.
As usual split into continents/regions.
Europe - Nice! I like what I see. About 2.5 Gb/s to London is no surprise (with Leaseweb) but very nice anyway. Solidly over 1 Gb/s to NL also is nice, and suggests that connectivity all over Europe is nice as well. And indeed it is, not a single target shows less than 200 Mb/s, except for (CH and) Spain.
Asia / Oceania - Overall decent, and not a single secondary target failed, nice!. Modulo some weird corners like the India, Mumbai target significantly slower than SGP (with is further away) and also Hongkong, and hardly faster than the China targets. But still, overall decent results with e.g. Ozzyland about 45 Mb/s.
Africa - Also quite decent, not high-end but really decent.
America - It shows that Leaseweb is routing via the UK; uncommonly good results to both the east coast as well as to the west coast. I mean, slightly above 80 Mb/s to both west coast targets isn't something I see often. Also note that, not the speed per se, but the ratio of the speed to NYC to the speed to Chicago looks much more reasonable than quite a few other VPSs.
TL;DR/Verdict: I would say that this UK VPS is the one that I'd like most (from my european perspective), if there weren't the Amsterdam VPS from HostDZire which also is based on Leaseweb connectivity. Re the network results, both the UK VPS and the AMS VPS are comparable, modulo a few weaker or stronger targets, and anyway damn good enough, but the HostDzire AMS VPS is significantly stronger wrt processor and memory, as well as in particular disk speed.
Also, NL not only is a EU member but closer to the heart of Europe, so i prefer it somewhat over a UK location.
That said, if your main concern is good cross-Atlantic connectivity along with still really good connectivity within Europe, the UK VPS might be the more attractive one.
Finally, a big THANK YOU! to @HostDzire who not only made this multi- (actually all) location review possible but who deserves - and gets - kudos from me for actually wanting me to find and identify any weak spots his VPS products and locations might have. Sorry, I have to "disappoint" you, HostDZire, I found almost none and those were very minor. *g
Btw, I have plans for my benchmarking and reviews and got myself one of HostDZire's own India VPSs which I intend to use as a kind of reference point in Asia, next to my Contabo SGP VPS which while certainly not great anymore still should be damn good enough for what intend to use it for. With HostDzire's India VPS traffic volume is a bit tight (but not too tight I think) and with the Contabo VPS bandwidth a quite tight (200 Mb/s) but then I've yet to see more than even 100 Mb/s in Asia ...
Thank you so much again for doing this
Its will be helpful for users to decide based on this.
Now they can see this and expect similar peformance, because yabs realy doesn't help alot. yabs provide good enough benchmark but its not detailed review so user might end-up not liking the vps later.
But when they see and read this, they will know what type of performance they will get.
So thank you so much for doing this.
You're welcome. But yabs IMO is next to worthless and largely marketing BS, especially wrt connectivity.
I recently had a look again at the yabs shell code and now also at the iperf3 (C) source code - and I found my attribution "marketing" BS" fully confirmend. For example, iperf3 works with "slots" that is, the target stack is fully available, and btw. those slots are not always available which boils down to potentially multiple waiting times for a free slot. And then it pumps as much through the connection as possible.
Don't get me wrong, that may indeed be useful - but for an entirely different purpose. For a large organization e.g. it may be useful to know how much traffic they can pump through a connection, or wavelength or fiber. But hardly for an average John Doe VPS user. For him/her it's important to know what real world performance they can achieve and without waiting for available slots ...
The mere fact that so many iperf3 servers, almost all of which are "sponsored" by carriers or providers, should make one think. At the same time it still isn't that easy to find good connectivity test files over a fast connection.
Besides, a single benchmark is next to worthless and like a lottery game. A real benchmark should be based on at least two or three dozen runs over (preferably multiple) 24 hr periods.
But I'll admit that I'm, a bit tired of my real world numbers being considerably lower than the marketing optimized high iperf3 numbers. So I've decided to make some changes; currently I'm working on http and https capability as well as on a compromise between iperf3 rainbow unicorn numbers and my modest single http connection numbers. And a few more, some of them major, changes and functionality are coming as well.
The test suites available for us mortal ?
If you mean mine, kind of yes. Explanation: version one had the source code publicly available. Result: lots of yada yada but hardly a handful of downloads. Next someone complained about me not using git. At that point I stopped making the source code available. But I gave a the binary (for linux) as well as docu available to a few users here who asked politely. And for one user I even created a Windows executable.
For the upcoming version 3 with considerably extended functionality and features I havent decided yet but quite likely will at least provide a decent docu and maybe even the binaries publicly.
Thank you if you make it available. An optimize benchmark test suites for us mortal for daily average resources usage.
Yet another HostDZire / Leaseweb location benchmark and review!
Good Lord, @HostDZire really is a nice guy. After I had discovered the CA promo VPS I asked him whether I could get access for a couple of days and expected a polite, even friendly, but still a "enough now, hmmm" response. Nope, he just said "Sure, no problem" and an hour or two later I found the CA VPS in my panel.
Thank you very much HostDZire! You successfully turned me into a HostDZire fan *g (that won't keep me away from staying honest, though, even blunt when justified).
So, as usual I'll start with sysinfo, processor, and memory.
No surprises and none expected, give or take basically the same as the WDC VPS, the "standard" Leaseweb VPS. So, really nice!
Let's look at the disk (probably very similar again).
Hmm, even significantly faster than the WDC NVMe. Not massively but give or take +10% write speed and +10% IOps is significant. But anyway, even the WDC disk is in solid performance territory, so me happy.
Finally the connectivity. If any then I'd expect differences here and that was my main point of interest.
Normally I'd do the usual splitting up into continents/regions, but I don't. The reason is simple: the results are very similar to the WDC VPS. Slightly better here, slightly worse there, but similar - and decent! - enough to not do the full walk. Maybe noteworthy: The WDC reaches slightly more targets. I guess the only way to get better connectivity within Murrica would be a more central location like e.g. Dallas (as experienced with the amazing @RIYAD Dallas VPS).
Rather I'll disclose why I was interested in this Leaseweb location: I wondered whether the Canada VPS would possibly be significantly better than my WDC reference VPS (for whatever miraculous reason ...). Answer: It is a fine NA location but the differences are small enough to stick to my WDC VPS, especially when I got it at a significantly lower price (back when it did cost only $24/yr which IIRC is almost half of the CA VPS's price) and also because over all I like my WDC VPS a bit more.
TL;DR/verdict: yet another "boringly" nice - and cheap with the promo! - HostDZire / Leaseweb VPS.
After doing quite many HostDZire and Leaseweb VPSs I think I can say that HostDZire really earned to be among my favourite providers. Yes, it may look a bit boring to see more or less the same or very similar results but (a) those results are very decent, and (b) in my books that's a plus because it means that Leaseweb quite consistently uses the same quite decent hardware in all locations and the differences mainly are in the location itself. Or in other words: when you buy a Leaseweb VPS you know what you get. Including connectivity in the sense of "quite good" everywhere.
That said, neither their hardware nor their connectivity is highest end but neither is their price, and both are clearly in what one might call solid business/professional territory. Yes, one can find faster hardware (like Ryzen and high-end NVMes) and one can find even better connectivity (like Hybula) but not or extremely rarely at a Leaseweb like price.
I very recently tested another provider's US promo VPS and I also have a Contabo VPS in both Murrica and Singapore; the last time I benchmarked them they were significantly worse than my Contabo VPS in Germany, which was upgraded and now is actually really decent (even in the top-3 list of one category in my "the bestest" shootout a while ago) but again, the last time I looked neither the St. Louis nor the SGP VPS had been upgraded. But I'm currently doing a benchmark run on both.
So, I guess my HostDZire India VPS ( their own, not Leaseweb) and the WDC Leaseweb VPS will be my reference for at least some time. Once they are upgraded (if they will be) I might add the 2 Contabo VPS into the mix. Europe is no problem, there the @Hybula Ryzen VPS was, is, and stays my reference.
Now, when a benchmarker who tested many, many VPSs from many providers, from super-cheap to kind of expensive, comes to the conclusion that a certain provider's products are the right choice for reference systems that's a major compliment I think. And a compliment HostDZire (and Leaseweb) earned.
@jsg
Thank you so much again
Yes, a bandwidth test is supposed to pump as much bandwidth as possible. That's its purpose. Iperf isn't perfect, far from it. I've filed bug reports years ago when dealing with bugs. But iperf maintainers don't throw whiny hissy fits when their design limits or coding bugs are pointed out.
You're not doing any real world tests, yourself. You trailed off without making a point about the slots. The slot wait time has no relevance to the to results.
That its expensive and require major bandwidth and players to run them so they're not overloaded.
I would argue testing just 24 hour periods isn't much different than yabs. You need multiple times a day, and weeks to arrive at reliable conclusions.
Then run yabs more than once. Seems self explanatory. But you realize that data is only useful if you graph the results over time, rather than average with no times.
If you're going to spend the time doing reviews on a regular basis, just put all your results in a spreadsheet, graph the results and release that data. Then actual comparisons can be done between hosts, locations, resources, etc.
If you're making a high bandwidth connection, you're not doing that over http (the hint is in the name). Nor have you made any sort of data available to prove its a real world test (your test is synthetic, you really have no idea what you're talking about).
Unless you're saying iperf doesn't count data correctly, you also don't know the meaning of "rainbow unicorn" numbers.
He used to put it out online. An earlier version, at least. Then when people would report getting nonsensical results (I provided detailed test results from Oracle I/O limited VPS and Cloud at Cost's next to unusable servers), he cried and whined and pulled the release so NOBODY can reproduce or validate his app. My phone is on low battery right now, but I'm sure that thread is before Oct 2020 if anyone wants to search for it.
I still have it if anyone wanted it. I'm away for another week but can post it somewhere next week if anyone actually wanted to test with it (spoiler alert, barely anyone did back then).
I've been private messaged by providers who tell me they know his numbers are nonsense, but they don't speak out for PR reasons.
Thank you for a detailed information. I think you have the point. We need multiple tests.
There's no information there, just nonsense, obviously wrong assertions, and bashing.
As for "we need multiple tests" I agree, that would be useful, but so far I know of none (other than hobbyist scripts and/or marketing tools).
Btw, his "been private messaged by providers who tell me they know his numbers are nonsense, but they don't speak out for PR reasons." statement obviously is BS and a lie too. (a) what problems ("PR reasons") would providers be afraid of when it's bloody obvious that even a nobody can consistently attack me and bash me?. And (b) provable and meanwhile even publicly known fact is that I've often been expressly asked to benchmark one or multiple products by providers, be it as a service for them (not published) or be it as a published review here.
Also btw. I've never deleted the code and/or binaries I've made available, it's still out there on the "Yandex drive" site. I simply didn't make the source available anymore (for newer versions).
Frankly, it's puzzling why he even still is tolerated here after many, many incidents of him spreading lies, attacking other users personally, commonly using vulgar language, aso, asf.
Let’s focus on the reproducibility. I’m not sure why you decided to stop. if everyone can “yabs” with your tools that would be a lot beneficial?
You can thank a (very few) real assholes for that. In fact I'm not even shure that I'll continue at all to provide that service to the LET community.
You see, it all started because I wanted real world performance numbers for myself, tried a few tools available and found the situation unsatisfying; those tools were not satisfactory and/or cumbersome and/or basically nonsensical (in my view). That's why I wrote my benchmark program in the first place. It was to scratch my itch and satisfy my curiosity. Publishing my results was just an extra on top because there was a desire for performance information.
So, I can simply go back to where I was after finishing my code and just scratch my itch and satisfy my curiosity.
Plus LET increasingly has become and continues to become even more a "only views count" shit hole anyway, sorry.
Yeah since you decided of what you want. There is nothing I can discuss anymore.
About LET becoming shit hole, afaik only HN can still strive with a lot of good discussions.
Reproducibility and standardisation is key to any serious benchmark. If you can't reproduce it yourself and make comparable assessments then the measurement itself is pointless.
LET,
What is my social responsibility to a provider no longer here who started a DM with me along the lines, "just between us, bro"?
Last active January 2022 and no longer in business?