oles@ovh.net
03-04-2013, 10:32 AM
Hello,
So what's New at OVH?
Lot of projects are being finished and being put into production. Here's a recap of what we have been doing with our days.
Datacentres:
------------
The RBX5 DC, with a capacity of 14,000 servers will soon be full, probably sometime in March. We're preparing RBX6 for the end of April. Then later in 2015. Infact we have bought a lot of new buildings in Roubaix Valley so we will be able to build RBX7 RBX8 RBX9 RBX10 .. but that's in two years. This is the time it takes to bring the EDF (electricity company) power supply of 20KV to the new address.
This is to avoid completely saturating RBX, we started the 2nd SBG DC : SBG4. It was put into production there 3 weeks ago and RBX5 is already working for PL, IT and DE clients. Also, the SBG2 DC is under construction and is expected in the 3rd Quarter 2013. In the meantime, we're finishing securing the site in fibre optics and increasing the electrical power through the work of GC in the Strasbourg port. SBG2 will really show its power by the end of 2013 thanks to the shortest network shortest path to IT that will build on our dark optical fibre.
To following on from RBX, we're planning GRA in Gravelines. The GRA1 DC will start in March. We are currently moving optical fibre to GC and connected on two new networks: Paris / London (bypassing Roubaix) and Gravelines / Brussels. The DC will have 3 outputs to 3 directions: London, Brussels and Paris, and will be connected to BHS and RBX directly through secure loops.
As you'll understand, March and April will be busy very busy months. We're doing everything to continue to deliver in 1hr, but between RBX5 which is full, SBG4 which is started GRA1 and RBX6 that will start, it may possibly be a few days, max 2 weeks of uncertainty, as we are running out of space in our DC. There is a risk, even if we're managing it. So if you have projects for March / April, you will need to order in advance to ensure they're not affected.
regarding Americia, we're selling more servers from our BHS DC. Because success is exponential, we started the construction of the 2nd DC, BHS2 whose capacity is 10,000 servers. Following that BHS3/BHS4 will be built following the RBX5 model, which allows us to delpoy 2000 servers in just 2 weeks. BHS3/BHS4 will also be powered, as in Europe by 400V instead of 480V like BHS1/BHS2. Infact, we have managed to standardise and validate European voltage in North America
We've started researching the construction of 3 DC's in the US: You can not be No. 1 in the world without being connected to the No. 1 market where growth rates are the most important. This is a project that should take us less than 2 years.
NETWORK:
-------
We have completed the capacity expansion of our USA network. Currently we carry 20% of our traffic from Europe through this new network which is already very well connected with American, South American and Asian operators. Concerning our DC in BHS: we have a little less than 500Gbps of capacity output and actual traffic exceeded 30Gbps and it's rising.
In Europe, we have completed the updates to our network along distance between RBX /LDN/ AMS. We have changed the technology (still on Infinera) and we're using 100G coherent to move our DC circuits between RBX and major points of presence in Europe. Next week, we resume work on AMS/BRU/RBX to complete the first loop and secure all optical circuits. This week, we'll update the Cisco ASR9 for RBX, Paris and SBG.
This will allow us to add capacity in our network in Paris but also fix bugs on netflow that prevent us from properly handling attacks.
Speaking of attacks, we're testing solutions from Arbor Networks, which can detect an attack, send it to mitigation device and neutralise it. We've also been testing for some months Tilera, which filters attacks in real time. The 2 of them have both given good results. We are in the process of choosing what solution we will retain to implement a comprehensive solution for each DC. The goal is to treat attacks of 200Gbps and more.
Coming back to the network, after the RBX/LDN/AMS/BRU/RBX loop, we will work on the RBX/FRA/SBG/PAR/RBX loop with the goal of securing all circuits. Why secure the circuits?
When there's an optical fibre failure (plant machinery that tears the fibre), the 10G circuit between 2 routers is cut, unless it is protected on the optical level by another way, When the fibre is cut, the circuit remains UP by equipment which automatically use another optical path. So there is no break in service. This type of network is more simple and less expensive to build in 100G coherent.
We should sign the optical fibres between SBG and Milan, in a few weeks and then implement this network in September. In parallel, we're looking for a location of a DC in Germany ..
CDN:
----
The BETA is going pretty well. We've fixedx bugs and we can already see more than 1.5Gbps of traffic on our CDN. Thank you to the BETA testers for their feedback which allows us to adjust the settings.
We're thinking of starting the first dedicated CDN offer in the next few days at €9.99 p/m for a number of unlimited domains with an Anycast IP.
In parallel, we have planned to start the DNS service based on the same infrastructure. Low Latency DNS aka LL DNS will cost €1 p/a per domain.
VPS:
----
This offer had a great start. We are very happy because we think its right for 100k p/a VM offers.
However, we have some concerns internally for the first deliveries. We're missing filer resources and we found a bug in the vCloud quite late .. so to summarise, we made some mistakes.
The new VPS Cloud are properly delivered and we have finished catching up with backorders. End of the week the first Cloud VPS will be all set.
The manager should be out in a few days. At that time, we will release the GAMMA offer which be finally bug free.
pCI 2013
--------
pCI 2013 is the private version of VPS with LB, VPN, FW and HD IO guarantees, multi site and by the hour. We're thinking of doing the BETA at the end of March.
hubic:
------
The infrastructure is stable and it's good. We are still testing different clients under iPhone, iPad, Android, Windows and WepApp. There are still a few days of work before making them public.
In short, as we have said all along, we want to give you an offer that works.
Customers who have trusted us and have paid a subscription, will have a 2nd year free. All this happens at the same time.
Shared hosting & GeoCache
---------------
Work on shared hosting now is well advanced. The main problem is due to the fact that the tools we developed to monitor the activity of certain sites were not taking into account the actual use of CPU / RAM for each site. We therefore found that a lot of big sites were using less than 40% CPU/RAM infrastructure capacity. As we approach 7000 HG servers, 40% is still 3000 HG servers. The latest work consists on replacing font scripts (the famous okillerd) to share resources equitably among all customers. It uses BigData and MapReduce to collect resource usage statistics on clusters. This allows us to see problems more quickly. We believe that these problems are still felt by about 5% of sites. Work continues to reach 0% then we will update to Hosting 2013 offers..
To accelerate hosting, we will integrate CDN by default, internally called GeoCaa which has made up 75% of our investments in this period. The rest is automatically financed via actuators that do not require the dividends at the end of each year. This will allow us to sustain our growth in France and Europe with your orders.
I'll stop there. It's a very long and heavy read to digest.
In short, we're working hard every day
Regards
Octave
So what's New at OVH?
Lot of projects are being finished and being put into production. Here's a recap of what we have been doing with our days.
Datacentres:
------------
The RBX5 DC, with a capacity of 14,000 servers will soon be full, probably sometime in March. We're preparing RBX6 for the end of April. Then later in 2015. Infact we have bought a lot of new buildings in Roubaix Valley so we will be able to build RBX7 RBX8 RBX9 RBX10 .. but that's in two years. This is the time it takes to bring the EDF (electricity company) power supply of 20KV to the new address.
This is to avoid completely saturating RBX, we started the 2nd SBG DC : SBG4. It was put into production there 3 weeks ago and RBX5 is already working for PL, IT and DE clients. Also, the SBG2 DC is under construction and is expected in the 3rd Quarter 2013. In the meantime, we're finishing securing the site in fibre optics and increasing the electrical power through the work of GC in the Strasbourg port. SBG2 will really show its power by the end of 2013 thanks to the shortest network shortest path to IT that will build on our dark optical fibre.
To following on from RBX, we're planning GRA in Gravelines. The GRA1 DC will start in March. We are currently moving optical fibre to GC and connected on two new networks: Paris / London (bypassing Roubaix) and Gravelines / Brussels. The DC will have 3 outputs to 3 directions: London, Brussels and Paris, and will be connected to BHS and RBX directly through secure loops.
As you'll understand, March and April will be busy very busy months. We're doing everything to continue to deliver in 1hr, but between RBX5 which is full, SBG4 which is started GRA1 and RBX6 that will start, it may possibly be a few days, max 2 weeks of uncertainty, as we are running out of space in our DC. There is a risk, even if we're managing it. So if you have projects for March / April, you will need to order in advance to ensure they're not affected.
regarding Americia, we're selling more servers from our BHS DC. Because success is exponential, we started the construction of the 2nd DC, BHS2 whose capacity is 10,000 servers. Following that BHS3/BHS4 will be built following the RBX5 model, which allows us to delpoy 2000 servers in just 2 weeks. BHS3/BHS4 will also be powered, as in Europe by 400V instead of 480V like BHS1/BHS2. Infact, we have managed to standardise and validate European voltage in North America
We've started researching the construction of 3 DC's in the US: You can not be No. 1 in the world without being connected to the No. 1 market where growth rates are the most important. This is a project that should take us less than 2 years.
NETWORK:
-------
We have completed the capacity expansion of our USA network. Currently we carry 20% of our traffic from Europe through this new network which is already very well connected with American, South American and Asian operators. Concerning our DC in BHS: we have a little less than 500Gbps of capacity output and actual traffic exceeded 30Gbps and it's rising.
In Europe, we have completed the updates to our network along distance between RBX /LDN/ AMS. We have changed the technology (still on Infinera) and we're using 100G coherent to move our DC circuits between RBX and major points of presence in Europe. Next week, we resume work on AMS/BRU/RBX to complete the first loop and secure all optical circuits. This week, we'll update the Cisco ASR9 for RBX, Paris and SBG.
This will allow us to add capacity in our network in Paris but also fix bugs on netflow that prevent us from properly handling attacks.
Speaking of attacks, we're testing solutions from Arbor Networks, which can detect an attack, send it to mitigation device and neutralise it. We've also been testing for some months Tilera, which filters attacks in real time. The 2 of them have both given good results. We are in the process of choosing what solution we will retain to implement a comprehensive solution for each DC. The goal is to treat attacks of 200Gbps and more.
Coming back to the network, after the RBX/LDN/AMS/BRU/RBX loop, we will work on the RBX/FRA/SBG/PAR/RBX loop with the goal of securing all circuits. Why secure the circuits?
When there's an optical fibre failure (plant machinery that tears the fibre), the 10G circuit between 2 routers is cut, unless it is protected on the optical level by another way, When the fibre is cut, the circuit remains UP by equipment which automatically use another optical path. So there is no break in service. This type of network is more simple and less expensive to build in 100G coherent.
We should sign the optical fibres between SBG and Milan, in a few weeks and then implement this network in September. In parallel, we're looking for a location of a DC in Germany ..
CDN:
----
The BETA is going pretty well. We've fixedx bugs and we can already see more than 1.5Gbps of traffic on our CDN. Thank you to the BETA testers for their feedback which allows us to adjust the settings.
We're thinking of starting the first dedicated CDN offer in the next few days at €9.99 p/m for a number of unlimited domains with an Anycast IP.
In parallel, we have planned to start the DNS service based on the same infrastructure. Low Latency DNS aka LL DNS will cost €1 p/a per domain.
VPS:
----
This offer had a great start. We are very happy because we think its right for 100k p/a VM offers.
However, we have some concerns internally for the first deliveries. We're missing filer resources and we found a bug in the vCloud quite late .. so to summarise, we made some mistakes.
The new VPS Cloud are properly delivered and we have finished catching up with backorders. End of the week the first Cloud VPS will be all set.
The manager should be out in a few days. At that time, we will release the GAMMA offer which be finally bug free.
pCI 2013
--------
pCI 2013 is the private version of VPS with LB, VPN, FW and HD IO guarantees, multi site and by the hour. We're thinking of doing the BETA at the end of March.
hubic:
------
The infrastructure is stable and it's good. We are still testing different clients under iPhone, iPad, Android, Windows and WepApp. There are still a few days of work before making them public.
In short, as we have said all along, we want to give you an offer that works.
Customers who have trusted us and have paid a subscription, will have a 2nd year free. All this happens at the same time.
Shared hosting & GeoCache
---------------
Work on shared hosting now is well advanced. The main problem is due to the fact that the tools we developed to monitor the activity of certain sites were not taking into account the actual use of CPU / RAM for each site. We therefore found that a lot of big sites were using less than 40% CPU/RAM infrastructure capacity. As we approach 7000 HG servers, 40% is still 3000 HG servers. The latest work consists on replacing font scripts (the famous okillerd) to share resources equitably among all customers. It uses BigData and MapReduce to collect resource usage statistics on clusters. This allows us to see problems more quickly. We believe that these problems are still felt by about 5% of sites. Work continues to reach 0% then we will update to Hosting 2013 offers..
To accelerate hosting, we will integrate CDN by default, internally called GeoCaa which has made up 75% of our investments in this period. The rest is automatically financed via actuators that do not require the dividends at the end of each year. This will allow us to sustain our growth in France and Europe with your orders.
I'll stop there. It's a very long and heavy read to digest.
In short, we're working hard every day
Regards
Octave