Srcasm

So, it was a grueling trial for us with the two vendors that we decided to look at, Riverbed with their Steelhead appliances and Juniper with their WXC boxes.  But we are finished.  If you want to cheat, you can skip to the end of this blog post and find out who we decided to go with — if you don’t know already — or you can read through all of the “junk” that I have to say about each product and what it had to offer us and you.

Juniper - The trouble child of the group…
I’ll begin with the Juniper box.  The WXC appliance and I got started on the wrong foot.  We decided to try to use their multi-pathing feature and our network design simply would not support it without a lot of work.  We spent days trying to figure out a way to connect the two routers and two VPN concentrators behind the box (on the WAN side) for redundant pathing with no luck.  We decided to go ahead without this test and figured that we would look into it later.

After we got over this hurdle, the WXC was simple to setup.  It provided us great value-add features such as QOS and reporting that was exactly what we needed.  The downside to the setup of the Juniper was that we had to specify subnets that were in front of the box (on the LAN side) and then also choose who we wanted to accelerate traffic with.  This would be great in some environments and control is key but in a meshed environment like we are moving to (MPLS), this was a headache.  Figuring out which subnets sat where and making sure that we didn’t miss any was a pain.   Other than these few issues that I ran into, the WXC was a great box.  It seemed to work well and saved us quite a bit of bandwidth of the trial.  Below is a screen capture of the numbers that we saved.  I was pleasantly impressed with this solution.

Riverbed - It was love at first site, sort of…
Riverbed was my choice from the beginning.  All of the reports that I had read and studies that I had gone over said that Riverbed was the solution for us.  Gartner puts them at the top-right of their 4-square rating system which was great for us.  During the sales pitch, Riverbed came out and showed us an in-house demo of the system.  They give you a stop watch and let you time the transfers and wait for your amazed look on your face and a blank check in hand.  We didn’t really have either since we knew what we were getting into.  We decided to delve deeper into this product and to bring a demo in line in a few of our sites.  We loaded up the gear, two for Corporate, one for Worcester, UK and one for Honoeye Falls, NY and set out on our journey.

After the setup of the first Riverbed box took a mere 7 minutes to boot up, configure and reboot to save the configuration changes and make sure it was running properly, I was impressed.  There were no tunnels to setup, no subnets to configure and when placed in line there was almost no downtime involved.  The beautiful part of this setup came when the UK put their box in line a few days later.  Since they’re 5 hours ahead of us here on the east coast, they put theirs in line about 0700 on a Thursday morning.  That was 0200 here on the east coast.  For the next 5 hours, they had no idea but they were compressing data, acceleration traffic and making their users lives easier.  When I came into the office, I got a call saying that the WAN had mysteriously improved exponentially.  I examined the setup and realized that not only had they put the box in line but they had booted it up.  This was great news — since it worked well — but it could have been a disaster.  Lucky for us, the Steelheads are smart enough to understand when a box is at the opposing end and they simply “make a connection”.  Below is a screen shot from the Riverbed demo showing just how well it was doing during its short life at Southco.

If you haven’t yet, you can read through the previous posts on the blog about our trials throughout this product selection process but we made our decision.  And, drum roll please…  If you didn’t figure it out by now, we decided on the Riverbed solution.  The pricing came out similar in both cases and Riverbed had an additional offering, the mobile client.  I don’t know if anyone understands how excited I am for the mobile client but I feel that it will make our users lives 100 times better while working remotely.  I have seen the client in action, I tested it out myself, and it is just as good as purchasing a $6000 appliance but it runs hidden on a computer and “just works”.

That’s all for this trial and decision.  Feel free to contact me with any questions or comments you might have and I’d be happy to give you a hand with any information you might need to better your WAN environment at your organization.

The Riverbed Experience - Week 1
Within the first day we, as well as our users, noticed a huge difference in WAN speed.  The users were saying that they had never seen such a fast connection back to Corporate and we were seeing upwards of 4-5 times WAN data reduction.  Since I haven’t talked about the technical workings of the Steelhead appliance yet, this seems like a great time to go over it.  The Steelhead offers three types of acceleration.  Data reduction, TCP acceleration and application acceleration.   Data reduction is the process of breaking down traffic to the bit level and caching it either on the box on a hard drive or in RAM.  This allows the device to send traffic over the WAN link with tags instead of whole data.  Second, the TCP acceleration portion helps to overcome the window sizing issues that I spoke of above.  It changes the way that TCP talks by tunneling the traffic over the WAN link.  Finally, the application acceleration piece is what Riverbed has in the bag.  It understands how applications such as MAPI, HTTP, HTTPS, and CIFS work among other protocols and accelerates these applications by sending data across that it knows that it will need.  This of it as a “type-ahead” sort of feature that you have on your cell phone.  As soon as you start to open an email, it might pre-send the commands necessary to reply to that email just in case you might want to in the near future and if you don’t it’s still compressing and accelerating the other data.

The Steelhead did an amazing job in our tests so far.  It helped the VOIP traffic a bit by not starving the line of so much bandwidth and it was able to send our Lotus Notes and CIFS traffic across the pipe at an amazing rate.  This was a sure winner in our book at this point and we knew our users and statistics would agree.

Juniper In Action - Week 1
The Juniper box works in much the same way as the Riverbed appliance.  It was able to accelerate data by using three different levels of acceleration.  It would compress data, it would accelerate TCP and it would do application acceleration.  The one thing that Juniper had over the Steelhead in the actual data it could handle was that it could work with UDP data as well.  This could be very helpful for companies that do a lot of video or voice traffic over their WAN.  We found it cause a larger headache for us though because we had not gone through the QOS setups on the devices.

The WXC worked very well though.  It gave us about the same increase in bandwidth as the Riverbed appliance and it provided an interesting additional feature.  Juniper calls it multipathing.  The idea with a multipath setup is that you can use multiple connections (such as a primary WAN connection and a DSL backup connection) to send data to the remote end at the same time.  This did not need any routing changes as it would simply tunnel the data over each connection.  This could help if you needed additional bandwidth or if you’d like to send your less important data down a different pipe (think backup data or email).  This feature does come with a price as it now causes you to create two tunnels or more to each location that you are connecting to but if additional bandwidth is important, this is a large win for Juniper.

Overall so far, both boxes have performed as we expected and in some cases even better than expected.  The setup of the Riverbed box was still simpler but the Juniper box had a nice feature that we felt would help Southco.  Well, enough for this post.  Next time I’ll go into our conclusion on this project.  Stay tuned.

The Riverbed Experience - Day 1
The install of the Riverbed hardware was unbelievably simple. Within 10 minutes I had the first box setup (with Riverbed’s help) and installed in the rack. We pre-configured the other three boxes and send them out to the specific sites.

The Juniper Experience - Day 1
Juniper’s WXC boxes had a fairly simple setup process as well. The downside to the Juniper setup was while there was a wizard, there are around 23 steps to set it up. These steps were simple but were still unneccesary screens to click through.  In addition, the screens were not as intuitive as the Steelhead’s screens.  In Junper’s defense, the WXC was able to offer a much better selection of QOS settings than the Riverbed appliance.  This could be very helpful in today’s networks especially if you are planning an MPLS meshed network soon.

So I’ve been working on a project at Southco to help speed up our WAN connections throughout the globe. Since we have 24+ sites around the world there is no doubt that there would be latency and bandwidth constraint problems. This is where this project came in. In the world of networking there are many pieces that can slow down a connection across the WAN link. First, there could be a physical barrier to overcome (such as getting data from Philadelphia, PA to Shi Yan, Shenzhen, China) which includes latency and jitter in the line — this is nearly unavoidable. Another major issue that we face today is the problem of chatty protocols and TCP causing slowdowns. TCP has window sizing which continuously shrinks exponentially each and every time a packet is lost in the process. Protocols such as CIFS and HTTP are extremly talkative and require many round trips to send just a small amount of data back and forth. Finally there are bandwidth constraints that play into slow-downs. Each day we send more and more data over the WAN links, just like every other company out there, and can’t afford the time or resources to continuously grow these network lines.

When I began to look at how to tackle this project I setup a list of items that we had to fix and then started reading. I read reports from vendors of WAN acceleration products, Gartner reports, independent research groups and hit the forums that I belong to to find out what other people were using. I found out two things very quickly — 1) This was not going to be an inexpensive fix in the short run, and 2) There was an abundant amount of offerings out there and they all had their advantages and disadvantages to each. After reading for many weeks on this subject I finally decided to try out two vendors’ products. The first was Riverbed’s Steelhead Appliances and the second was Juniper’s WXC.

So the next task was to begin the demos. We started out talking to Juniper and Riverbed and explained what our network architecture looked like. That went smooth with both vendors. We got in the four demo boxes from each vendor to place in our Corporate location and three other locations and began the demo of the Riverbed appliance.