Choice of ISP on CityFibre for low latency (London area)

I have a bit of a niche requirement - because I regularly play live with a band online we need very low ping times and jitter (even lower than most gamers would need) - anything more than single figures makes a noticeable difference to us. I am currently on an FTTC connection with Plusnet, and the ping times to servers in London are around 8ms (bbc.co.uk around 7ms), so very usable - I'm using a modified router with OpenWRT firmware including SQM to minimise jitter during other internet use in the house.

Our area went live with CityFibre in May - we currently have a choice of 7 ISPs including Vodafone, TalkTalk and Giganet. Yayzi has recently been added, which at present is the cheapest. The area is also largely covered by Openreach FTTP, but not our street.

Our Plusnet contract runs out in November, at which point I would like to move over to FTTP, and the CityFibre offers seem attractive. My question is: I understand that ping times will be routing dependent so inevitably the choice of ISP will be a bit of a lottery. I don't think any providers offer a money back guarantee based on latency. What can I do to maximise my chances of getting minimum ping times? I'm doing the obvious research, and hope to ask around the neighbours when there's enough CityFibre take-up, but is there anything else I can do? And does anyone in the forum have experience of this?

Many thanks

Being based in / near London definitely helps. You *should* be blue to match or better the ping times on your FTTC connection. I’d expect better than 5ms.

I’d check with the neighbours though and see what they’re achieving.

For what its worth, I'm on Giganet 900 (CF) in Maidenhead since April (so not too far from London)

Pings to BBC.co.uk (light/no load) are 5ms (wired LAN)

Pinging uk.www.bbc.co.uk.pri.bbc.co.uk [212.58.233.253] with 32 bytes of data:
Reply from 212.58.233.253: bytes=32 time=5ms TTL=54
Reply from 212.58.233.253: bytes=32 time=5ms TTL=54
Reply from 212.58.233.253: bytes=32 time=5ms TTL=54
Reply from 212.58.233.253: bytes=32 time=5ms TTL=54

BQM here shows this (I run multiple web/ftp/mail servers on this which may explain some of the yellow spikes throughout the day)


Under full load 900mbit down or up (according to ookla and others) this can go up to ~10ms but then I guess you wont be maxing it out with a few A/V streams!.
I'm using OPNSense on a fairly complex setup on a Proctli VP2410 (because my older Draytek router struggled to get anywhere near full speed esp with IPv6 enabled (the router that Giganet supplied had no problems mind, just not as configurable is I'd like).

Good feedback Phil.

Under full load 900mbit down or up (according to ookla and others) this can go up to ~10ms but then I guess you wont be maxing it out with a few A/V streams!.

That's pretty darn decent actually for unloaded vs loaded ping.

Yeh I was quite impressed myself especially coming from a shocking experience with Virgin "buisness" who were far far worse than what I got with FTTC prior (epecially jitter)...

Just for fits and giggles, I just ran another ping whilst running a speed test (delivering 900+ in both directions)

Ping statistics for 212.58.233.253:
Packets: Sent = 45, Received = 45, Lost = 0 (0% loss),
Approximate round trip times in milli-seconds:
Minimum = 4ms, Maximum = 30ms, Average = 12ms

Interesting, during maxed out download, it averages 11ms but during maxed out upload, it was a little slower, around 20ms - that may be down to my router, cant be sure and for my usage scenarios, not something to be concerned with)

Much is down to your own machine (especially if running a speed test & simultaneous ping - try a separate machine for each) and also your own router.

However it all looks excellent.

Probably getting a bit off topic, but your right I hadn't considered that.
As you suggest, running the ping directly on the router whilst speed test on a different box gave completely different results - well during the upload test at least.

During download (on the server)
64 bytes from 212.58.237.253: icmp_seq=11 ttl=56 time=10.737 ms
64 bytes from 212.58.237.253: icmp_seq=12 ttl=56 time=12.192 ms
64 bytes from 212.58.237.253: icmp_seq=13 ttl=56 time=11.716 ms
64 bytes from 212.58.237.253: icmp_seq=14 ttl=56 time=12.056 ms
64 bytes from 212.58.237.253: icmp_seq=15 ttl=56 time=11.068 ms
64 bytes from 212.58.237.253: icmp_seq=16 ttl=56 time=12.035 ms
64 bytes from 212.58.237.253: icmp_seq=17 ttl=56 time=10.264 ms
64 bytes from 212.58.237.253: icmp_seq=18 ttl=56 time=11.630 ms
64 bytes from 212.58.237.253: icmp_seq=19 ttl=56 time=10.353 ms
64 bytes from 212.58.237.253: icmp_seq=20 ttl=56 time=11.775 ms
64 bytes from 212.58.237.253: icmp_seq=21 ttl=56 time=12.453 ms
64 bytes from 212.58.237.253: icmp_seq=22 ttl=56 time=10.611 ms
64 bytes from 212.58.237.253: icmp_seq=23 ttl=56 time=11.899 ms
64 bytes from 212.58.237.253: icmp_seq=24 ttl=56 time=11.505 ms
64 bytes from 212.58.237.253: icmp_seq=25 ttl=56 time=11.717 ms

And during the upload test
64 bytes from 212.58.237.253: icmp_seq=27 ttl=56 time=4.327 ms
64 bytes from 212.58.237.253: icmp_seq=28 ttl=56 time=4.957 ms
64 bytes from 212.58.237.253: icmp_seq=29 ttl=56 time=5.109 ms
64 bytes from 212.58.237.253: icmp_seq=30 ttl=56 time=4.601 ms
64 bytes from 212.58.237.253: icmp_seq=31 ttl=56 time=4.532 ms
64 bytes from 212.58.237.253: icmp_seq=32 ttl=56 time=4.536 ms
64 bytes from 212.58.237.253: icmp_seq=33 ttl=56 time=4.662 ms
64 bytes from 212.58.237.253: icmp_seq=34 ttl=56 time=4.943 ms
64 bytes from 212.58.237.253: icmp_seq=35 ttl=56 time=5.382 ms
64 bytes from 212.58.237.253: icmp_seq=36 ttl=56 time=4.569 ms
64 bytes from 212.58.237.253: icmp_seq=37 ttl=56 time=4.559 ms
64 bytes from 212.58.237.253: icmp_seq=38 ttl=56 time=4.347 ms
64 bytes from 212.58.237.253: icmp_seq=39 ttl=56 time=4.731 ms

Much better

--- uk.www.bbc.co.uk.pri.bbc.co.uk ping statistics ---
46 packets transmitted, 46 packets received, 0.0% packet loss
round-trip min/avg/max/stddev = 4.327/7.059/12.453/3.103 ms

👍

Thanks for the information, everyone - particularly to Phil for his ping results from Maidenhead. I'm minded to go with Giganet as they are cheaper than Vodafone and Talktalk and don't increase their prices in contract, but if anyone has results from a different ISP (particularly Yayzi as they are much cheaper) I'd be very interested to hear.

We're actually just down the road from Phil in Reading so should hopefully get similar results here. For the online band we use a Microsoft Azure cloud server as one of the band members gets free access through work, and we all get ping times in the range 7-9ms. SQM on our router limits the on-load ping times to within a couple of ms of off load.

Edited by Stargazer99 (Wed 10-Aug-22 10:01:42)

Ummm?
Yayzi is Blackpool based and their website is down at the moment. Has been for at least 15 minutes.

The right-hand column of a Google search I just did on my iPad has a summary, the website link, and opening hours 8am to 6pm except closed on Sundays. Facebook seems to think closing at 5pm not 6pm.

Company registered December 2019, no financial info yet except fixed assets of £72,000. Registered by Liam Mulryan as the sole shareholder.

David Bailey bought at least 50% shareholding 3 August 2020. Two more directors appointed during 2021. David Bailey sold his shareholding back to the company and resigned as a director 6 November 2021.

Twenty Trustpilot reviews, for what that's worth, not impressive. The oldest 8 all 5 stars, grim thereafter including one three days ago.

I can't see the prices of course until the site is back up. I'm not impressed.

Edit: Site is back up, will look through it later.

Edited by pluralist (Wed 10-Aug-22 12:12:52)

The good thing with Giganet (which persuaded me) was that they offer 1Mo, 12Mo and 24Mo contracts - I went with 12 (1Mo didn't get the 3 months free offer) but it means you can change sooner if there is a problem. Their pre-sales support was also really helpful anserwing lots of tecy questions (above and way beyond the usual home broadband expectations) that others were not so keen to answer. Although I only had a choice between Zen and Giganet initially, I asked a few others too). Haven't needed support since installation.

Having read around the forums recently, I'm glad I didn't go with Zen as they seem to have some latency issues if/when you get switch to their Manchester gateway.

Good luck with your investigations!

Yes, it does seem that Giganet is the one to beat - the only reason I would choose another ISP is if they are just as good but cheaper. I would probably take the 1-year contract for the discount. There are only two of us in the household and we don't even have a TV that is UHD compatible, so for us the Giganet 150Mbps deal at £32 per month would be ample.

Yayzi indeed doesn't look very promising from their recent reviews. The other ISP that has just turned up that has very competitive pricing is Hyperfibre, which I gather is operated by Marstons Telecom, an offshoot of the brewing company - they are listed on the CityFibre website as a provider in our are although when I check on their own website it says they aren't live yet.

So yes, probably Giganet unless there is a compelling reason otherwise. thanks to everyone for their help

BUT (yes, I'm having an 'and another thing...' moment!)
I have several of my own domains for which my email address is connected to my Google account. Originally Google just allowed me to send email from them via my account but stopped this in 2016 - they still allow me to associate an email address with my Google account but outgoing mail must be routed via an ISP's SMTP server. Currently with Plusnet this is no problem as an email address comes as part of the package - I give Google the settings for Plusnet's SMTP server and it just works. But I can't find any such information on Giganet - maybe this isn't part of their package. So what should I do to keep the customised email addresses (preferably without paying extra)?

Edited by Stargazer99 (Fri 12-Aug-22 17:06:50)

If you are receiving incoming mail through those domains via gmail, usually the host(s) of those domains will already be providing SMTP and IMAP (or POP3).Some though still provide a simple forwarding service, which might be what you are using.

In the first case which I have, (TsoHost and 1 & 1), with several domains I use Thunderbird on my laptop with gmail separate. On my phone I use Bluemail for the domains plus gmail again separate.

On my daily use domains I have usually the lowest level mail package. Thunderbird and Bluemail handle them fine using the host's servers.

If you only use forwarding, with no email package, AIUI you can still pay Plusnet to let you use their SMTP.

If any of your domains has a mail package you could simply use that instead of Plusnet. If none have, I think I would prefer to get it on one of them, paying them instead of Plusnet.

I don't *think* Giganet have outgoing SMTP relay but I'm not sure as I never thought to ask them (perhaps that would be good to know) - Maybe its worth dropping the customer services an email and ask them before you make your mind up. They were very responsive to my complex questions.

I host my own email for several domains (on my Giganet connection but I have backup connections/servers) but all outgoing email is routed out via a couple of VPS's that I own (in the "cloud") as Giganet does not offer things like RDNS on the residential services. They do have business services but they are completely different, leased line type of thing and not the same infrastructure (read incredibly expensive, not designed for the likes of me!!)

Edited by philg (Fri 12-Aug-22 13:36:39)

I've emailed Giganet - they replied promptly but don't have an external facing SMTP server I can use. At present I forward all email for my domains to my Gmail account at the domain registrar's control panel (my domains are registered with Fasthosts) - the web element of my personal domains is just a forwarder to my LinkedIn profile so they aren't hosted anywhere else. I do have separate hosting for various websites I manage for third parties - this uses a cheap US hosting company and does have mailservers included in the package.

So potentially I could use my hosting company's mailserver instead (will probably have to point my domains towards the hosting company's nameservers then add a forwarder to LinkedIn for web traffic there instead). I could then set up an empty mailbox for each address I want to use, for SMTP only, and hopefully Google will be happy using the hosting company's SMTP server. A bit more involved but hopefully feasible.

In reply to a post by Stargazer99:

So potentially I could use my hosting company's mailserver instead

They may enable outbound without having to handle inbound, worth asking.

Or look at authsmtp.com which works really well, but can be costly for high volume.

I tried changing the DNS to point the web traffic to my cheap hosting and added an HTML forwarder, which works, and have also added a mailbox for one of my custom domain email addresses to the hosting using cPanel. I can then add the address to my Google account with SMTP via my hosting and incoming mail forwarding still redirecting via the registrar's servers, and everything works. I can also use Thunderbird with the same SMTP settings to send mail out.

Only problem now is that Google's incoming filters aren't keen on the arrangement so when I send mail via my hosting provider's server to my Gmail account as a test it gets flagged up as suspicious, whereas it didn't seem to mind me using the Plusnet relay. So the setup works but with reservations - particularly as my wife's business also uses the same arrangement.

I could pay a small amount for a Fasthosts mailbox for my wife's mail, but I'm not sure whether that would support SMTP relay for my other addresses. There seem to be a lot of stand-alone SMTP relay providers out there but they're all rather expensive for anything more than about 100 emails per month. Any recommendations?

Edited by Stargazer99 (Sun 14-Aug-22 23:44:06)

In reply to a post by Stargazer99:

Any recommendations?

Not really, the problem you found is the main one, "reputation" of the sending setup is critical when you need to email anyone on other systems (e.g. Google, Microsoft, Yahoo, or the corporate versions of these).

The best way is to buy a mailbox service from a reputable provider, e.g. Google or Microsoft, that includes receiving and sending. e.g. Microsoft's Exchange Online is only £3 + VAT a month, but with a 12 month committment, for a mailbox. You need to have a domain name and point the MX records at the MS service. Google has similar products.

https://www.microsoft.com/en-gb/microsoft-365/exchan...

Post deleted by Tazz_uk

In Berkshire here on Giganet over Cityfibre. Had it about 6 months now and not dropped one 4ms ping and excellent gaming latency and always max speed on speed tests:

ping bbc.co.uk

Pinging bbc.co.uk [2a04:4e42:200::81] with 32 bytes of data:
Reply from 2a04:4e42:200::81: time=4ms
Reply from 2a04:4e42:200::81: time=4ms
Reply from 2a04:4e42:200::81: time=4ms
Reply from 2a04:4e42:200::81: time=4ms

Its so good I have recommended 2 people now.

Good news - I think I've found a way to fix the reputation of my outgoing email (and have just found out a load of stuff about SPF [Sender Policy Framework] as part of the bargain!). I'm using the same configuration that was previously flagging up a warning in Gmail but have added an SPF text record to the DNS configuration for my domain which effectively indicates that the third party hosting company's relay server address is one that I as the domain owner have validated for sending mail. The email header has gone from SPF neutral (unknown) to SPF pass (validated by domain owner) and the warnings in Gmail have disappeared. If this also works for my wife's business email (which is currently using the Plusnet SMTP relay) it looks like I should be able to migrate to Giganet without any concerns.

Tazz_UK - those ping times look great, very reassuring. The online band I play in uses Jamulus, an open-source client/server based system for playing music together online - in our band we all achieve ping times to an Azure cloud-based server of about 7-9ms and an overall round trip delay including audio interface overheads etc. of around 25ms (that equates to playing our instruments about 25 feet apart on a real stage, which is realistic). Another 4-5 milliseconds off that and a more inherently reliable FTTP connection should work brilliantly.

Edited by Stargazer99 (Mon 15-Aug-22 13:46:31)

And... Giganet over CityFibre was finally connected up here today. I'm getting ping times to London servers of 5-6ms. Could always do with that extra millisecond off that, but it's fine for my requirements and 2-3ms better than Openreach FTTC. Thanks to everyone for their recommendations

1 <1 ms <1 ms <1 ms dsldevice.lan [192.168.1.1]
2 5 ms 5 ms 5 ms thn-bn-2.giga.net.uk [37.48.224.7]
3 5 ms 5 ms 5 ms 217-168-248-28.m12solutions.net [217.168.248.28]
4 27 ms 11 ms 24 ms lonap.as13335.net [5.57.81.75]
5 6 ms 6 ms 6 ms 141.101.71.2
6 6 ms 6 ms 6 ms one.one.one.one [1.1.1.1]

Am I being dumb here, but isn't pinging one of the mainstream DNS servers to check latency a bit pointless? Surely that says nothing at all about latency to a "production" website/game server?

It's the latency to a/the target site that matters.

CloudFlare DNS… doesn’t get much more “production” than this. This website uses their CDN

In reply to a post by Pheasant:

CloudFlare DNS… doesn’t get much more “production” than this. This website uses their CDN

Depends if the DNS uses the CDN (unlikely) or uses multi-cast (more likely). Applications are likely to be fronted by a content network / distributed cache, but eventually your client application needs to talk to the actual server making decisions.

Yeah my point was really that their DNS seems a fair enough target comparator for checking latency.

Edit:

I see a couple of relevany posts have been made while I messed around drafting and altering this post.
================================================================

That's my point . But don't forget, pinging 1.1.1.1 doesn't get past the first Cloudfare router that is set to respond. It doesn't go anywhere useful.

Surely it would make more sense to be pinging the site they use?

In reply to a post by Stargazer99:

The online band I play in uses Jamulus, an open-source client/server based system for playing music together online - in our band we all achieve ping times to an Azure cloud-based server of about 7-9ms .... Another 4-5 milliseconds off that and a more inherently reliable FTTP connection should work brilliantly.

Witness also the ongoing discontent in the Zen forum here about users near London being randomly connected to Zen gateways in Manchester and from there to Telehouse servers in London. And similarly Zen users in the north-west randomly connecting to gateways in London instead of Manchester. Using 1.1.1.1 wouldn't help them. They still go through their connection to Zen.

Unlikely to happen with Giganet/CityFibre, but who knows? We'll soon find out, assuming the OP reports back.

As for CDN and this site, whilst I'm aware Cloudfare does something for it, I don't believe the whole of all tbb databases are cached on it. The OP's band playing streaming through Jomulus and Azure is by definition not cached either.

Similarly Microsoft and Apple may use it for software releases and updates, but their complete databases? YouTube? The whole of the BBC website content? A fair bit of topical stuff will be cached of course.

Edited by pluralist (Fri 30-Sep-22 13:39:54)

In reply to a post by Pheasant:

Yeah my point was really that their DNS seems a fair enough target comparator for checking latency.

See my previous post. He had the comparator suggested by James.

Edit: But of course he couldn't

Edited by pluralist (Fri 30-Sep-22 13:44:23)

Pinging his Plusnet DNS server would probably have given the same as 1.1.1.1

No. Not necessarily - check out the trace route above.

A bit of context to that traceroute to 1.1.1.1, and a small confession - it's actually a cross-post from a different forum where I was having a similar simultaneous discussion, where I was replying to a poster who illustrated his Giganet/CityFibre latency using that same address (and as it turned out we had exactly the same routing and almost identical ping times), so it's your OP being a bit lazy with copy/paste I think of 1.1.1.1 as being illustrative of the best (or near best) achievable to a real-world server, with most others being worse of course. From my experience it also gives a pretty good idea of the ping time I should get to a London-based Jamulus cloud server, which is what I'm most interested in. Anyway, the route seems to go directly via Telehouse in London, which is what I need for low latency - hopefully with Giganet the route is unlikely to deviate far.

I can now get the following ping times to Jamulus servers:
Fastest ping 5ms (London)
Typical London-based cloud servers 6-8ms
Nearest server outside the UK 8ms (Groningen, Netherlands)
Germany 15-28ms
East coast USA 73-80ms.

With my equipment, to get the real round trip delay time (or at least what Jamulus indicates, which I believe is accurate) you need to add 12-17ms for audio processing etc to the ping time depending on the quality of the route and the server, so the best servers in London would give you about 17-25ms. In real world playing terms at the speed of sound this corresponds to the musicians being about 5-8 metres apart, which is realistic for a band playing together. Much more than 10-12 metres and you start having difficulty keeping time.

In reply to a post by Pheasant:

No. Not necessarily - check out the trace route above.

I assume you mean this one:

1 <1 ms <1 ms <1 ms dsldevice.lan [192.168.1.1]
2 5 ms 5 ms 5 ms thn-bn-2.giga.net.uk [37.48.224.7]
3 5 ms 5 ms 5 ms 217-168-248-28.m12solutions.net [217.168.248.28]
4 27 ms 11 ms 24 ms lonap.as13335.net [5.57.81.75]
5 6 ms 6 ms 6 ms 141.101.71.2
6 6 ms 6 ms 6 ms one.one.one.one [1.1.1.1]

That is basically doing what I said. Remember, he is using a vISP.

Step 2 his connection goes to them;
Step 3 forwarded to the actual provider which is giganet;
Step 4 from them to Lonap which is a Layer 2 exchange in London;
Step 5 is in fact Cloudfare;
Step 6 is 1.1.1.1

That a pretty complex routing (all through high performance kit of course).

I stand by my opinion that if he had pinged his Plusnet DNS server his latency would have probably been the same. It may in fact have been 5ms or even 4ms as it would be almost certainly have been Step 2. We'd need a Reading Plusnet user to check (I think the OP said he is near Reading).

Plus I still believe it's irrelevant for anyone to ping a DNS server.

Thanks for that Stargazer. Looks like all is good, which is what you and the rest of us basically expected. Enjoy .

The discussion over the value of pinging 1.1.1.1 is interesting,at least to me, but now irrelevant to your case. It may not be to someone elsewhere in the country.

Yep there's only one traceroute posted in the thread

Its actually a very simple traceroute too; a minimal number of hops and vitally using 1.1.1.1 as a reference makes sense as its only 2 hops removed from LONAP so no convoluted routes to complicate matters.

To your earlier comment - I'm not actually sure that the OP could ping Plusnet DNS boxes from outside their (that is Plusnet) network ie from a Giganet connection. Whereas Cloudflare DNS is reachable globally and therefore is just as relevant a latency comparator as pinging say bbc.co.uk for example. They are also reasonably quick for a public provider.

I'm not sure what you mean by vISP?

Giganet are an ISP in their own right. AS numbers and the whole nine yards. There is nothing virtual bout them. If you're thinking that they run service over CityFibre FTTP - well CF are much like Openreach in that they are not an ISP - they provide a transparent pipe from the user to their FEX / point of presence / national backbone and from the ISP must backhaul and get their user traffic to / from the internet.

Our Plusnet FTTC service hasn't been disconnected yet, so here are the corresponding routes when I change over the cable to the Plusnet router. 2ms faster to Telehouse, which is reflected in the ping times I get from Jamulus servers:

>tracert 212.159.6.9

Tracing route to cdns01.plus.net [212.159.6.9]
over a maximum of 30 hops:

1 <1 ms <1 ms <1 ms OpenWrt.lan [192.168.1.253]
2 39 ms 14 ms 9 ms 250.core.plus.net [195.166.130.250]
3 7 ms 7 ms 7 ms 84.93.253.87
4 7 ms 7 ms 6 ms cdns01.plus.net [212.159.6.9]

Trace complete.

>tracert 1.1.1.1

Tracing route to one.one.one.one [1.1.1.1]
over a maximum of 30 hops:

1 <1 ms <1 ms 1 ms OpenWrt.lan [192.168.1.253]
2 6 ms 6 ms 6 ms 250.core.plus.net [195.166.130.250]
3 7 ms 6 ms 7 ms 84.93.253.83
4 7 ms 7 ms 7 ms core1-BE1.colindale.ukcore.bt.net [195.99.125.132]
5 7 ms 7 ms 7 ms peer7-et-7-0-1.telehouse.ukcore.bt.net [109.159.252.92]
6 8 ms 8 ms 8 ms 109.159.253.95
7 10 ms 8 ms 11 ms 172.70.160.4
8 7 ms 7 ms 7 ms one.one.one.one [1.1.1.1]

Trace complete.

In reply to a post by Pheasant:

To your earlier comment - I'm not actually sure that the OP could ping Plusnet DNS boxes from outside their (that is Plusnet) network ie from a Giganet connection. Whereas Cloudflare DNS is reachable globally and therefore is just as relevant a latency comparator as pinging say bbc.co.uk for example. They are also reasonably quick for a public provider.

The Plusnet dns servers that I use are 212.159.6.9 and 212.159.6.10
I can ping them from a Gigaclear connection so I don't think that you have to be using a Plusnet connection.

Cool. So fact of the matter is that your new FTTP service from Giganet (over CityFibre) is on average several milliseconds faster than your existing Plusnet FTTC service (to the same point).

Hope all goes well with the band. Enjoy!