New Member

Question

Share your WAN Optimization experiences here

Forum|Forum|15 years ago
May 5, 2011
20 replies
11356 views

So I finally took the time to try WAN Optimization between two FortiGate' s. Both are 60C' s with 32 GB Class 10 SD cards installed, and 29 GB provisioned for WAN Optimization storage space. There are Windows servers at each site with SMB file shares. I did a rule set on each site to cache traffic requested by client PCs to the servers on each side. I made two rules: one for port 445, which I think was enough, but just in case I also made one for 49152-65535, which is the random port range Server 2008/Vista+ use in conjunction with port 445. The results were fairly immediate and somewhat impressive. Downloading a 30 MB file will take 10 mins the first time, and only ~1 minute the second time - peaking at 2 MB/s file transfer versus maybe 40 KB/s for a non-cached file. However, I do notice increased latency versus before, so I' m reluctant to roll out optimization to the server to server traffic (the servers seem to use 445 for a lot of communication) and to traffic outside of CIFS/SMB. I do think this has great potential but the Fortinet docs and knowledge base are lacking on examples for site-to-site traffic optimization (e.g., generic TCP optimization mentioned in the " Inside FortiOS" sheet but totally absent in any examples). If people are interested in my rule sets that I got to work I' d be happy to share them, and conversely I' d love to hear from others that have deployed WAN Optimization and where it is benefiting them.

Rp44580.jpg

Maik

New Member

Hi veechee I did my wanopt test last year and expected more benefits than i really got. I also noticed the " added delays" . what network latency do you have between your two sites? could you please try the following for me? 1) create an office document (word, excel etc) of any size (e.g.1mb) 2) transfer it twice from local to the remote site -> do you see the caching effect? 3) then on the local site, open the office document and close it again (notice, no " save" was done). 4) transfer it again -> do you still see the caching effect? thanks

emnoc

New Member

Has anybody seen WAN-opt bake-offs between the other players just how well does fortinet stack up to cisco, riverbed, sliverpeak,etc....? I' ve personally have not tried any wan-opt outside of the lab, but have use the others mention above. And all has some benefits over fortinet in their approach, methods, reporting, statistics gathering and overall performance & bandwidth savings.

kinderu28

New Member

Hi guys, I' m new on this forum so please be gentle. I' m doing my thesis on wan optimization using Fortinet equipments(Fortigate 51B, version 4.0 MR2). I' ve managed to see how traffic has been optimized (according to the reduction rate graph in the monitor menu) for all protocols,except 1.MAPI. TCP,CIFS,FTP,HTTP have been successfully optimized by the equipment except for the MAPI protocol.After 3 days trying to deploy Exchange server(AD,DNS,Outlook) i finally got some MAPI traffic by sending emails(with or without attachments).I' ve managed to send across the link about 30-50MB of MAPI traffic, but there was no wan optimization.Reduction rate 0%. My test consisted on sending a mail with an attachment and then basically downloading the same attachment several times via Outlook.I' ve used transparent mode and byte caching int the wan rule for the MAPI protocol and used a peer to peer configuration .In my opinion this should' ve resulted in some kind of reduction. Have any of you managed to obtain any reduction for the MAPI protocol?Some feedback or any kind information regarding this aspect would be very useful. Thanks a lot

simonorch

Explorer

Same here, i' ve got MAPI hitting the wan opt rule but i' m not getting any reduction.

veecheeAuthor

New Member

Re: MAPI In Outlook 2007 and 2010 the option exists to encrypt traffic between the Exchange server and the client. Is that enabled? This is just a guess, but I think that the encryption would remove the ability for the Fortigate to optimize any traffic. kinderu28: What are you experiences with latency when using WAN Optimization on all those protocols?

Maik

New Member

before you invest to much time in WanOpt & MAPI, I recommend to open a support ticket and wait for their response

kinderu28

New Member

I' m using Exchange 2003 and Outlook 2003. @veechee: I haven' t quite used o monitoring tool for measuring latency(haven' t found one yet:) ),although i noticed that http performs quite well if data is cached.It' s a lot faster when downloading a file that was previously cached.However FTP and CIFS don' t seem that faster,but i have to make more tests like: - 10.000 files of 100 Kbytes - 10 files of 100Mbytes - 1 file of 1Gbyte I will get back to you shortly(1 or 2 day max) PS:i achieved 98%reduction rate for TCP traffic using jperf. PS2:i notice that memory usage is quite high(50-60%) and sometimes sends the unit into conserve mode. PS3:My 2 FortiGates are directly connected via the wan1 interface.I was thinking of creating a routing loop between (3-4 hops) to really simulate a WAN environment

kinderu28

New Member

@Maik: for some reason i can' t open a ticket or submit my product. Actually,it is not my product, it belongs to my University, and it could be for teaching/testing purpose only.I don' t know.Both 51-B have the same serial number:FG50BH3G0*******.I' m not sure if i can give away this information,that' s why i haven' t given the full serial.

veecheeAuthor

New Member

kinderu28, Based solely on my personal observations, WAN Optimization on CIFS and the generic TCP optimization, increases the latency, so for example, pulling up a folder is slower, but once I initiate a file transfer if it' s cached it goes faster overall. I' d definitely recommend you try to simulate slower speed links AND latency, as these bare the things WAN Optimization is supposed to help overcome, and they do not exist in a direct connection situation.

romanr

New Member

Hi, As Outlook uses local caching of the exchange mailboxes by default in versions 2007 and 2010, Mapi support is somehow a legacy thing... without local caching of the clients you should get some good results with MAPI! br, Roman

Maik

New Member

roman are the " good results with MAPI" only a guess or did you actually test and see it by yourself?

romanr

New Member

Maik, I had some good resulsts with Mapi in lab testing only... But as the new Outlook version do cache the mailboxes very well (which also can get activated in 2003), the MAPI caching will in real life only be really helpful for public folders or so... if they get used... We had some very bad experiences though with CIFS when used with DFS in a 2008R2 network... sometimes the traffic between the FGTs maxed out without traffic from the clients... So we did not implement it in a real production environment... br, Roman

Maik

New Member

Then it looks like i need to update my WanOpt know how. It' s currently based on a support ticket information which became outdated with your test: " [...] MAPI protocol is currently supported for the purposes of identifying traffic. This is why you can observe that protocol is recognized, but is not actually optimized. [...] Current implementation only support general TCP acceleration for MAPI traffic, which may reduce latency for MAPI traffic through WAN. Other possible optimization such as byte-caching, compression, pre-fetching are not implemented." the MR3 release notes did not indicate a feature update in that direction, so I did not test again. back to lab

Show more replies

Sign up

Login with SSO

Login to the community

Login with SSO