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
Network Tests
Download speed from CacheFly: 62.8MB/s
Download speed from Coloat, Atlanta GA: 25.1MB/s
Download speed from Softlayer, Dallas, TX: 39.1MB/s
Download speed from Linode, Tokyo, JP: 12.2MB/s
Download speed from Leaseweb, Haarlem, NL: 51.0MB/s
Download speed from Softlayer, Singapore: 6.29MB/s
Download speed from Softlayer, Seattle, WA: 22.5MB/s
Download speed from Softlayer, San Jose, CA: 21.1MB/s
Download speed from Softlayer, Washington, DC: 37.9MB/s
Download speed from Dacentec, Lenoir, NC USA: 84.4MB/s
Disk IO Test
I/O speed : 19.8 MB/s
Well, I guess mine is okay lol.
I had opened an issue the other day about slow network performance. Weren't able to find the issue..... but I think this could be the culprit!
Just gotta figure out how to change the cache-settings within ESXi without having to reboot. Anyone knows perhaps? Can't find it in the VMWare KBs.
@linuxthefish - 100Mbps unmetered is available on certain systems.
Here's some performance comparisons:
@msg7086 - same test as you ran to same server - almost twice the download speed:
You misinterpreted my statement, I was being generous in my words. Without question Delimiter's network outperforms, more redundant and better scaled; thats why my comment was for performance/redundancy you choose Delimiter, for more storage you choose Dacentec.
Obviously you have to give something to compete, you're trying for the price game so you can't go like-for-like because people will say that the network slow, higher latency, worse connected. So you have to find a selling point which in your case was bundling 4TB of storage. Kudos to you for finding a niche.
But anyway, as you want to discuss networks:
We are not connected to the Telx's IX any longer, we connected for about a year but after we got direct peering with Amazon and couple of other networks there was no value to be there. We pick up the major networks at other exchanges or have paid peering with them.
This may change again in the future but it will require more interesting peers to be available on that IX. Its not a tier 1 IX, its mostly small regional players with a few larger content providers who get free ports to help drive participants.
As for network differences, you had an outage a few months back on 12th June, due to:
http://www.lowendtalk.com/discussion/comment/1112790/#Comment_1112790
'Charter lost a 100 gig bundle to Atlanta, it is still down. There was some routing instability failing over to Charlotte, but it seems all the issues have been resolved. Contact [email protected] if you are still having any issues.'
So what are you saying is that you have an unprotected (ie non-redundant) connection from Telx in Atlanta to your datacentre. So when Charter does maintenance, has an outage or whatever; all your connectivity is down with the exception of Telia who you seem to connect to in Charlotte.
If you had any redundancy on that connection you wouldn't have suffered such an outage, your router would simply have dropped the routes on the affected circuit and continued routing.
When I look at your routes, I see almost 100% of your traffic going via Telx, so in the event of a failure at Telx, a failure on your unprotected ethernet (?) service from Charter, you lose almost all your transit.
I am tracerouting from New York to give you the benefit of a Northern approach to your network. The assumption is that most decent networks would pass through NC on the way to GA rather than go up from GA to NC as a stub.
If I traceroute from Telia in New York to mirror.dacentec.com:
The Telia route in Charlotte is visible.
Traceroute from Level3 in New York to mirror.dacentec.com:
Handed over by Level 3 in Atlanta
Traceroute from Cogent in New York to mirror.dacentec.com:
Back through Level3 in Atlanta
Traceroute via PCCW
Again via Level3 in Atlanta
You have two transit providers, HE are not providing you transit but peering over Telx IX in Atlanta:
Level 3 in Atlanta backhauled over this unprotected Charter connection. HE peering in Atlanta which is over the same Charter connection.
Telia in Charlotte is unprotected.
Its very precarious.
Just had to send a ticket in... Ran their bench.sh on my dacentec server..... with an ssd.... getting this..
^ what the absolute hell.
@Foul - Thank you for proving my point!
Not happy with my network speeds either, lol.
Actually i have no idea why it's so slow on upload. I'm on optimum 75/35mbps and can only get 3MB/s download speed on ftp from my dacentec box. Way too slow compared to kimsufi which just saturates my port from the beginning to the end.
@Foul - well I am happy to help you move to our network
@msg7086 - I am going to PM you a download link on our network - would you try it and tell me what performance you get.
Here is mine dacentect server speedtest:
Retrieving speedtest.net configuration...
Retrieving speedtest.net server list...
Testing from Centrilogic (162.248.xx.xxx)...
Selecting best server based on latency...
Hosted by TELUS (Toronto, ON) [146.41 km]: 56.911 ms
Testing download speed........................................
Download: 48.34 Mbit/s
Testing upload speed..................................................
Upload: 36.41 Mbit/s
@shivoham - you can't test from NC to Canada, its not a fair test.
For comparison try the one that @msg7086 used:
speedtest --server 663
Impressive.
@msg7086 - so it saturated your broadband connection
Yup. So i have no idea about dacentec network. Maybe @shivoham can share something about that IP subnet. I heard that some of their subnets get slower connection.
China
My kimsufi servers seems to give better network speed then dacentec:
Retrieving speedtest.net configuration...
Retrieving speedtest.net server list...
Testing from OVH Hosting (192.95.xx.xxx)...
Selecting best server based on latency...
Hosted by Cubicule Studio (Montreal, QC) [1.98 km]: 3.165 ms
Testing download speed........................................
Download: 93.75 Mbit/s
Testing upload speed..................................................
Upload: 92.51 Mbit/s
Wow they give /48 IPv6 also!
From my dacentec
Hosted by Optimum Online (New York City, NY) [410.74 km]: 52.058 ms Testing download speed........................................ Download: 184.37 Mbit/s Testing upload speed.................................................. Upload: 79.44 Mbit/s
my speeds aint the best either.
Bristol, UK
Roubaix, France
Scranton, PA
Garden City, NY
Greenville, NC
All of those who are using speedtest.net's crap, I don't think you can judge network's speed from it, if you have 1ms connectivity to speedtest's server then most likely you'll have appr. full speed (~1000MB/s if speedtest's server has Gbit port). And if you've 100+ ms latency them most likely you won't get similar speeds.
Hope you guys have got my point.
@MarkTurner, I don't think you have a clear picture of our network. Our links are protected, and we are in Telx 125 Noth Myers, Charlotte as well as Atlanta. We have around a half dozen fiber providers landed in our telco room. Our link to Level 3 (Rose Lake, Charlotte, NC) should also be up soon.
For IP transit, we are in the middle of changing bandwidth, so taking down Tinet/Nlayer links while we get connections back to Cogent and add Level3, and the last two months we have been transitioning. Redundancy plus one has been maintained throughout the transition, although it has been less than optimal at some times.
We also have IP links in the building for gaming customers that aren't seen on the usual tools.
http://prntscr.com/7whmkv
Telx isn't just for Amazon connectivity. What about apple, cloudflare, facebook, google twitter and microsoft? Also the other hosting companies like OVH, hivelocity, softlayer and the mobile networks like sprint and tmobile.
http://bgp.he.net/AS7363#_peers
http://bgp.he.net/AS31863#_peers
@dacentec - That Charter backhaul from Atlanta certainly wasn't protected, thats why you had that meltdown during their network failure/maintenance.
If we're talking about the same place Telx Charlotte site its like the size of a toilet. The only national transit providers there are Level3, XO and TWTelecom (will leave shortly as Level3 and TW merge networks). So 2 transit providers!
If you have all these 'half dozen fiber providers landed in our telco room' then its even more shocking that you don't have more transit in place. There is no point having 6 fibre providers in place and you only have Telia in Charlotte.
Telx Atlanta - We pick up almost all of those networks elsewhere, some have no value to us or our customers. When you look at that IX, most of the participants are small regional operators. For us to maintain a connection to an IX, we need to pushing at least 5Gbps sustained otherwise its not worth the headache.
For us, even IX's like DECIX NY have far better potential (https://nyc.de-cix.net/customers-partners/customers/). The people behind are actively persuing the larger networks onto it, whereas Telx its just a way to sell another cross-connect.
If you hadn't have taken offense at my statement about your and our strengths, then this discussion wouldn't be happening.
The fact of it is that you're not in a primary transit area, you have lower network performance so you are compensating for that by bundling larger storage. Its not an insult, its just a statement of fact.
Servers in roughly NC/SC area
Honestly I don't understand why this is going on.
Has Dacentec a "premium" network? No, they didn't claim to either.
Do they have a decent network for the price? Yes, they do.
No need to hate. They also offer more traffic per server than Delimiter, for obvious reasons. Different offers, it's fine.
Also, the SpeedTest crap doesn't tell much.
If you go back to my original comment someone asked how does Delimiter/Dacentec compare, I answered it that Delimiter provides better network, Dacentec provides more storage. Nothing inflammatory. Dacentec then tried to upsell their network and present as something it clearly isn't and here we are.
It was not intended to start some Lowend Drama, just a polite observation.
@MarkTurner test their network http://lowend spirit.com €3/yr @AutoSnipe uses them.
Same thing using charter actually and got only 700KBs from them via sftp and they said it was my provider.
I checked Verizon fios I checked my vpn etc. Same thing. Kimsufi saturates my home connection as well.
Mark, Check Charlotte again, Level3, GTT, Tinet, Cogent, Telia and the rest are in Charlotte. We are connected to Telia, we used to be and are going to re-connect with Cogent. We are connecting to the Level3 gateway instead of going through the carrier house simply because it's easy because of where we are and the fiber partners we have.
I doubt google, apple, facebook, disney, bed bath and beyond, wipro and the rest would build datacenters in northwest north carolina if there was a connectivity issue.
Only for customers who prefer a Cogent/TW blend to a Level3/TeliaSonera blend
And as Sady pointed out, customers may have different results depending on who they are connected with. Dacentec understands that, and we back our service up with a 7 day no questions asked refund policy for accounts without spam or abuse.
I didn't say they weren't in Charlotte, I said they were not in Telx Charlotte. Please re-read what I wrote.
TW/Level3 = the same network, we now see all of Level3's routes as TW's primary transit routes. Additionally XO provides fantastic routes aggregating Level3, Telia, GTT and a few other Tier 1 providers.
The results in this thread alone show Delimiter's network outperforms yours by a significant magnitude.
Just look at the appalling performance in this thread, 700KB/s to Charter. Just OMG! The same customer got 8Mbps from us.
Another one above got some other pathetic performance on Optimum Online, same test with Delimiter and his connection was saturated.
We can keep at this all day but the numbers prove my point. Maybe we need to ask Delimiter to put a 2x2TB offer on the table and see how many customers jump ship? I've spoken to many of the customers who have churned from Delimiter to you and the overriding reason was lack of storage (2x2TB vs 500GB).
Network has never been a reason that Delimiter churned a customer to Dacentec.