Expected network bandwidth (Mbps) is missing for the B-series #8765
Comments
|
@Bartolomeus-649 thanks for the feedback. The B-series is variable, so it doesn't make sense to publish networking numbers that may not always be accurate. A similar request was also brought up on this doc before: |
|
Well, then you should put that in the documentation! You can't just not document it when you state that last column should contain the number of NICs and Expected network bandwidth. And even if it is variable, then one one to know inside which span will it vary? Because I guess it's not between 0 and infinity? I'm quite sure there are something to say about the networking of the B-series. |
|
@Bartolomeus-649 fair enough. @jonbeck7 can you take a look and consider adding some information? |
mimckitt
added
assigned-to-author
doc-enhancement
and removed
cxp
docs-experience
labels
|
@Bartolomeus-649 Thank you for the feedback. I took out the wording about bandwidth being covered in this article, since it isn't provided for all sizes in this grouping. I'll also have a talk with the owner of the B-series to see whether there is anything else we can add to this going forward. #please-close |
|
Ok, so this resulted in even less info on network performance being available. Why not just let the owner of the B-Series group write the documentation themselves? It would be much more efficient, any it would be in their interest to provide good documentation, if they, themselves would have to spend time reacting to user feedback on that documentation. Also, they are the only ones who know when the doc needs to change because some new feature och when something changes. |
|
@Bartolomeus-649 The owner of the B series is the one who determines what needs to be present in the doc then we have an author draft it. In the case of the B series you are not the first asking for this information. The design of the B series does not provide the user with a consistent level of network performance so providing specific numbers on them is not something we want to include. Even if it is a varying number those numbers could not be 100% precise and could lead to expectations that the B series itself cannot keep up with. In addition, the B series is designed for more CPU intensive workloads as that is what we are really scaling out depending on credits. If network bandwidth is an issue then the B series is likely not the best choice. If the B series product group decides this information is critical for users to understand and can think of a good way to present it we will update the content. Until then we have decided no to add that information but have made the doc more clear by removing information that would suggest it is something customers should be concerned with. We appreciate your suggestion and it has not gone unheard it is just at this point in time we cannot commit to providing accurate information about the B series networking bandwidth worthy of being documented. |
|
Ok, so what you are saying is that we can't even be sure to get network connectivity at all...the bandwidth could be zero bytes/s, and if we get network in the B-Series it could be limited and never go over 100 kbit/s or even less, all based on completely random and unknown factors? |
|
@Bartolomeus-649 obviously we would ensure you have network connectivity and for many cases it should be more than sufficient for the users needs. Again, these machines are not designed for top network bandwidth and if that is something you will need I would suggest another series. It is safe to assume the networking capabilities of the B series should be sufficient. In any case if you hit issues with a cap you could move to another series that has a guaranteed bandwidth expectations. We just do not have enough concrete numbers on what the bandwidth would be to document. Saying it could be between 0 and 10GB at any given point would cause just as many concerns if it was not a consistent bandwidth we could guarantee. Especially if an application requires a consistent 1000MPBS but at times it drops below and goes above would still be an issue. |
|
As there is NO information on the network capacity of the B-Series, it's impossible to know if you can use it for anything! Things such as link speed, any throteling going on, if when you have multiple NICs in the VM, are they mapped to their own physical NIC, ort can they share the same physical NIC? What is the network speed of the physical host? Is the issue with uncertain VM network performance related to the fact that all VMs on the host share the the physical NICS and there is no fixed (min/max) allocation set for how much a VM can use the physical NIC?, What about different types of network traffic? Any QoS in play that affect this? There is so much that can be said without promesing a specific throughput. It's like saying you have a "string", without saying how long it is, or it's thickness, or anything else about the string...perhaps it is a rope you could use to secure a huge ship, or perhaps it is a sewing thread...who knows? . |
|
I suppose it is a bit like an app service plan. You simply choose the amount of CPU and RAM you need and don't worry about the throughput per say. And these of course would also be used for websites and what not. The same situation would apply if you looked at AWS's T2 series which is similar to the burstable series we offer. The burst is more concerned about CPU rather than network performance. We are talking with the owner of the B series to see if there is any more information that we can add. I completely agree that more clarification could be added. But until we have confirmed there is nothing we are able to add to the document at this time. I would not want to add anything that is not 100% accurate. I will make sure to include your notes as well as the points you are bringing up are valid. Main thing I suppose at this point would be if you are concerned that the B series might not meet your needs then you should pick another series. Seeing as this series of machines are fairly new there is only more and detailed information we will be able to provide going forward. |
|
@cynthn please include me on any conversations with the B Series owner so I can be sure to voice any concerns on this doc :) |
|
@MicahMcKittrick-MSFT started a thread with someone from sizes and the folks that provide the networking stats :) |
|
@Bartolomeus-649 just an update, @cynthn and I have been working offline with the product groups who own the B series. We are not sure what information to add at this time but we are working to add additional information. It may not be the fastest turn around as it does require some research and better understanding of what we should and should not put out there. But rest assured we are working on it and do plan on adding more information as soon as we can. |
|
It must at least bursts to 1200 baud... no idea. |
|
is this an official statement from microsoft?
|
|
@marcurdy With minimal empirical testing (aka speedtest.net), I get around 800-1600Mbps down and 100-800Mbps up from a West US 2 B1ms. |
|
That's many times more then enough to choose B-series, but even a range is
nowhere to be found. Thanks.
…On Tue, Jul 10, 2018, 5:29 PM Timothy Elmer ***@***.***> wrote:
@marcurdy <https://github.com/marcurdy> With minimal empirical testing (*aka
speedtest.net <http://speedtest.net>*), I get around 800-1600Mbps down
and 100-800Mbps up from a West US 2 B1ms.
—
You are receiving this because you were mentioned.
Reply to this email directly, view it on GitHub
<#8765 (comment)>,
or mute the thread
<https://github.com/notifications/unsubscribe-auth/AFfdRIbPU3MLOqgES3Q1skk1RiDeVglDks5uFSrOgaJpZM4UC-JE>
.
|
|
cool, we now know more about the B-series than Microsoft does... |
|
To be fair to Microsoft, this range is huge, and likely very dependent on bandwidth usage at the providers, and which zone the instance lives in. I totally understand why they would be hesitant to publish numbers, but at least a low estimate would be nice. |
|
As mentioned, we are aware of the numbers but they do vary. So posting publicly can lead to false expectations. We are working internally to determine the best way to display this range. We appreciate all the feedback and rest assured it is being addressed. |
|
It won't lead to false expectations if you tell all the facts around what affect the network. |
|
@Bartolomeus-649 Maybe. Keep in mind that this is ONLY on my West US 2 instance. Other zones (or maybe other instances in the same zone) might be different. |
|
sure, but before we had ZERO bits of info on the network, Microsoft would only confirm that the B-Series indead had network cards, but that's it. |
|
As @telmer6 mentioned, that was just his zone. If you want to know for yourself in whatever zone I would suggest running a test as well until we can verify consistent speeds across all regions and make that public. All your feedback is forwarded to the product teams. Thanks. |
|
Just for more info, I've tested my tiny B1ms and get 1200 Mb/s down and 871 Mb/s up. (east us region) |
@MicahMcKittrick-MSFT, @cynthn You getting this? Still have not run the B-Series long enough to have average and median bandwidth, average and median latency and so on for the different versions of the B-series? |
|
Hello, One more interested in knowing the bandwidth of machines B. I'm a Microsoft Partner. How do I guarantee the bandwidth of B machines to my customers? We have made closed packages of virtual machines for my clients. How do I inform them about the bandwidth? It is a topic that should be seen by Microsoft. At least put a minimum and a maximum of bandwidth. Let's give our feedback in the following link: regards |
|
Been searching for B series network bandwidth and end up here... |
|
Is there any update on this? I am currently transitioning some of our small web servers to B series SKUs and whilst I can setup effective monitoring and alerting for things like:
... I am unable to setup effective monitoring and alerting for the VM network throughput. If at least the minimum and maximum network throughput range was provided, I could setup monitoring and alerting at the minimum range. I would like a way to understand if the B series SKU is causing a bottleneck on network throughput, but currently, I am unable to do so. |
|
Only what the response was earlier: "The B-series is variable, so it doesn't make sense to publish networking numbers that may not always be accurate. A similar request was also brought up on this doc before: |
Expected network bandwidth (Mbps) is missing for the B-series
Document Details
⚠ Do not edit this section. It is required for docs.microsoft.com ➟ GitHub issue linking.
The text was updated successfully, but these errors were encountered: